Pavilions have traumatized me about as much as any structure in the game.  Not only must we screw up our nice smooth loops with stub roads if we want to put pavilions in our housing blocks, but after the pavilions are installed the blasted entertainers don’t stay in the loops anyway, but go teleporting off onto the surrounding service roads.  I hope this post may help other players who, like me, want more control of their pavilions’ entertainers.  In particular, this post focuses on teleportation: the sudden appearance of a walker on a square not adjacent to the building that dispatched him (or her).  Teleportation has to qualify as a pretty specialized subject within the general area of walker studies, so readers should expect much of the material below to sound like complete gibberish unless they have studied the introduction to roaming walker behavior in “Ambulomancy” offered through a link given in another thread in this forum, “Predicting Roaming Walks”.
   Sometimes, the entertainers generated by pavilions really wind up in the weeds - completely off the road.  Although this post describes in general terms the conditions that produce such behavior, a full treatment of off-roading by entertainers and architects (and possibly others?) will have to wait for a separate post devoted to that subject.

Critical squares.  All three kinds of entertainers treat a pavilion as if it consisted of nothing but the 2x2 dance stage.  The locations of the music and juggle stages are ignored.  The walk-start and -finish squares of the dance stage (and therefore the entire pavilion) can be determined as Brugle describes for common walkers from 2x2 buildings in “Random walker start and finish points”.  Entertainers arriving from their schools as destination walkers are a bit trickier.  If the stage for his/her type of arriving walker is unoccupied, then the walker disappears on the walk-start square of the dance stage.  If a show is in progress on the stage for entertainers of the new arrival’s type, then the walker disappears on the walk-finish square of the dance stage.  The north square of the dance stage also serves as the “governing square” from which the locations of the four routing centers can be determined by the usual procedures outlined in Ambulomancy:  eight squares away in all four semicardinal directions.

Teleporting entertainers.  Of the three kinds of entertainers dispatched by pavilions, musicians and dancers teleport, but jugglers do not.  However, jugglers, as we will see, do play a pivotal role in facilitating teleportation.  Like the other entertainers in Pharaoh, pavilion-dispatched entertainers are medium walkers with 35-square default walks.  Like a bandstand (see "Bandstand walkers"), a pavilion has only a single quadramble which is shared by all of its walkers (entertainers and the labor recruiter).  Each roaming walker dispatched by the building is assigned the next formal direction in a single, clockwise cycle of walks.

Initiation of teleportation.  Starting with the first game of Pharaoh most builders ever played, we all noticed that a small amount of time passes between the instant a roaming walker who has just returned to his building disappears on his walk-finish square and reappears on his walk-start square to begin his next walk.  I call that interval the walker’s coffee break.  Teleportation from a pavilion will not happen unless two entertainers are simultaneously taking a coffee break.  Teleportation is not guaranteed to happen if a second entertainer begins his coffee break only an animation frame or two before a first entertainer was scheduled to end his break;  a certain amount of time (not yet precisely measured) together in the break room seems to be required to trigger teleportation, although as best I can tell it is not even as long as it takes a walker to advance a single square of travel.  In the discussions below, when time becomes important, I quantify time in terms of “squares of travel” (SoT), because I measured time using a modification of Jimhotep’s clock described in “Pharaoh Calender”

Jugglers as teleportation catalysts.  The game has a trick both to increase and to decrease the likelihood of simultaneous coffee breaks (and, therefore, teleportation):  changing coffee-break duration.  Long breaks increase the probability that two walkers will have overlapping breaks.  The principal way that the presence of jugglers stimulate teleportation is by increasing break length.  When the juggle stage is unoccupied, musicians and dancers take short breaks, usually lasting just two to four SoT.  While a show is in progress on the juggle stage, all the pavilion’s walkers take long, self-indulgent coffee breaks usually lasting 22 to 24 SoT.
   Keeping breaks short is not the game’s only method for inhibiting teleportation in the absence of jugglers.  If a musician and dancer arrive in fairly close succession (but not simultaneously) at a juggle-free pavilion, the first arriving walker gets yanked out of the break room and booted out onto the street as fast as possible.  The second walker is then detained for a fairly long (for a juggle-free pavilion) break of 10 to 12 SoT.  This difference in break time substantially increases the separation between the arrival times of these two entertainers when they return to the pavilion.
   Despite the algorithm’s efforts to decrease the likelihood of teleportation in the absence of jugglers, teleportation from pavilions does occur without jugglers.  For this to happen, however, a dancer and a musician returning from a walk must disappear on the walk-finish square of their pavilion at almost the same time.  A two to three animation frame difference in arrival time is close enough to simultaneity to permit teleportation, but a two to three SoT difference in arrival times is too great to allow teleportation.  Like a chemical reaction with a high activation energy, teleportation is accelerated by the presence of an appropriate catalyst (jugglers) but will proceed slowly without the catalyst.

Jump teams.  When two or three walkers have been on break together long enough for teleportation to happen, they form a “jump team”.   One walker in each jump team (hereafter called the “spotter” or non-jumper) does not teleport and all the other team members do.  If the jump team includes a juggler, he will be the spotter.  Otherwise, the musician spots for the dancer.
   All the members of a jump team seem to the player to appear simultaneously on their walk-start squares.  I have been unable to detect so much as a single animation frame of difference in their times of appearance.  Nevertheless, the algorithm clearly assigns them a formal order of appearance which I have never observed to vary:  Jugglers appear first, then musicians, and finally dancers.  The order of appearance does not depend upon the order of arrival of the team members at the pavilion (as they return from their previous walks); the order depends only on occupations.

Remote walk starts.  When the jump team is dispatched by the pavilion, the first team member (usually the juggler) appears on the pavilion’s usual walk-start square and is dispatched on a walk with whatever formal direction (e.g., NE) is the next one due for execution in the pavilion’s clockwise cycle of walks (quadramble).  Other members of the jump team are assigned later walks in the pavilion’s quadramble in the order of their (closely timed) appearances.  Thus, for a three-man jump team, if the juggler is assigned a NE walk, the musician will be given the SE walk, and the dancer will get the SW walk.  Unlike the juggler, who uses the pavilion’s usual walk-start square, a musician or dancer in a three-man team uses a non-standard, teleporting walk-start square: the walk-target of the previous walker to leave the pavilion.  In this example, the musician will use the walk-target of the juggler as her walk-start square, and the dancer will use the walk-target of the musician as a walk-start square.  The use of prior walk-targets as current walk-start squares certainly gives the appearance of teleportation, since the musician and dancer suddenly wink into existence up to 14 squares away from the north square of the dance stage.
   Although they begin on unusual squares, the walks of musicians and dancers following teleportation behave in many ways like the walks of other roamers in Pharaoh, allowing for differences in walk length (since entertainers are medium walkers and nobody else is except tax collectors).  Unfortunately, the other walkers in Pharaoh whom teleporters from pavilions most closely resemble are walkers with “bad habits” like tax collectors (who vanish at the end of some of their walks rather than returning, visibly to their walk-finish squares) and architects (who not only sometimes vanish, but on other occasions wander right off the roads).  Fortunately, the analytical tools needed to understand  and predict the behavior of pavilion roamers after teleportation are ones that apply to all roaming walkers (routing centers, road-search order, walk targets, destination- and random-mode portions of walks, etc.) and that have already been described in Ambulomancy ("Predicting roaming walks").  The remainder of this post explores the various kinds of roaming walks that entertainers from pavilions execute.

[This message has been edited by StephAmon (edited 08-23-2005 @ 05:20 PM).]

Grounded walks.  Grounded walks for spotters and teleporters follow all the usual rules.  If the shortest-road connected path between walk-start square (traditional or teleport-style) and walk target has length of 34 squares or less and contains no roadblock, the entertainer’s walk follows that path (in destination mode) and converts to random mode (for 35 more squares of travel) without reversing course on top of the walk target.

Illusory counter-clockwise rotation.  In the March 2005 edition of Ambulomancy ("Predicting Roaming Walks"), I stated that teleporting entertainers appeared on “landing” squares in a counterclockwise rotation.  This observation appeared to violate the rule applying to all other walkers that says their walks must progress in a clockwise rotation of formal directions.  The counterclockwise rotation is now revealed as an artifact.
   Observations of individual entertainers leaving pavilions from the usual walk-start point and two- and three-man jump teams have clearly revealed that all the roamers from a pavilion draw their walks from a single queue and the walks within that queue are arranged in a clockwise progression of formal directions.  Figure 1 provides the starting point for an example.  Even though the order is imperceptible to the player, the juggler is assigned his northeast walk first, then the musician gets the next (SE) walk, and finally the dancer begins a SW walk.  All three entertainers will return to the walk-finish square of the pavilion at more or less the same time - close enough to permit teleportation.  Therefore, they will all leave the pavilion on their next walks, again, as a three-man jump group.

Figure 1.  Critical squares for the grounded walks of a three-man jump team.  Pop-ups on the plazas and oddly colored road tiles specify their roles in governing teleportation.  Small statues mark routing centers (NW not shown).  The road to the NW is much longer than shown.  North lies toward upper left.

   The next walk available, after the targets shown in Fig.1 have been used, has a formal direction of NW, so the juggler (in the second round of teleportation) will appear first as the spotter for the group on the pavilion’s usual walk-start square, and begin a grounded walk with a NW target.  The musician will teleport to a remote walk-start square (the juggler’s target) for a grounded walk to a NE target (the same one shown in Fig. 1 as a yellow plaza).  At the same apparent time (but actually last of the three), the dancer will teleport to the musician’s walk target (now the yellow plaza square) and begin a grounded walk to a SE target (the ordinary plaza square in Fig. 1).  To the player following only the dancer, the teleportation-landing (or remote walk-start) squares appear to be used by the dancer in counterclockwise order (e.g., a SE landing in Fig. 1, then a NE landing for the second round of walks).  The same thing happens if the player watches only the musician.  When all four of the walks in the pavilion’s quadramble are grounded walks of similar length, this apparent rotation can continue indefinitely, as long as all the entertainers leave the pavilion as part of three-man jump groups.  Thus, I perceived a sequence of 270-degree clockwise rotations (for individual performers) as a series of 90-degree counterclockwise rotations.  Now, we know better.

Shorted walks.  The designers evidently changed the way the walker algorithm handles shorted walks for pavilion entertainers (and for architects) from the way those walks are described in Ambulomancy for the majority of walkers.  Instead of just two kinds of walks that can be produced either by roadblocks on, or by excessive lengths of, the shortest circuit, these walkers have at least three.  One of the three kinds of walks can send the roamers off road, and (as mentioned earlier), is still the subject of on-going research.  In this work, I content myself with approximately describing the conditions that can trigger off-roading.
   In the following discussion, I once again use the letter d for the length of a default walk: 35 squares for entertainers; 43 squares for architects.

Walks shorted only by blockage.  If the distance of the shortest circuit from a building’s walk-start square to the target of a walk is d - 1 squares or less and a roadblock lies on that path, walks of ordinary roamers short out for reasons that can only be attributed to blockage; without the roadblock, no shorting would have occurred.  The outward-bound halves of the walks followed by entertainers and architects under these circumstances look exactly like blocked shorts for ordinary walkers.   Unlike most walkers, pavilion entertainers and architects vanish at the end of the outward-bound half of their only-blockage-shorted walks, i.e., after d squares of travel from the walk start.
   When architects vanish at the end of a walk shorted only by blockage, they ghost home invisibly.  They do not show up on the usual game display, but their presence and movement can be detected by passing the footprint of a large building (as if the player was about to position that building) above the shortest path between the architect’s vanishing point and his office.  Ghosting architects travel off road and (probably) diagonally.  Their ghosts appear to travel in a series of steps rather than a smooth diagonal, but they arrive at their destinations much faster than they could if they were traveling only in NE-SW and NW-SE directions.
   In tests of pavilions with just one kind of entertainer, the roamers usually reappear on their next walks just two to four SoT after vanishing at the end of walks shorted only by blockage.  Thus, sufficient time was not available for the entertainer to have walked invisibly back to the pavilion.  When an entertainer, who is part of a jump team, disappears after a blocked-only short, too little time also passes in many cases for the walker to ghost back to the pavilion.  Also, no ghosts of pavilion entertainers have been detected between their vanishing points and their walk-finish points.  Thus, unlike architects, entertainers from pavilions do not appear to ghost.

Walks aborted by distance.  If a walk in a pavilion’s quadramble assigns an entertainer an unobstructed target lying 35 (d) squares or more from the roamer's walk-start (traditional or teleported), the entertainer’s walk does not short out by having its length reduced to d squares (like those of ordinary walkers).  Instead, the entertainer simply walks to the target and vanishes, thus appearing to abort the walk at its midpoint without returning to his walk finish.  If a roadblock lies further than d squares from the walk-start point along the shortest road-connected path to the target, the entertainer flouts the roadblock by passing through it.  If a road-block appears earlier, a confused walk is likely to be generated (See next section.).

Walks confused by blockage and distance.  If the shortest road-connected path to the entertainer’s target is further than d squares from the walk start (so aborting due to distance would be possible without a roadblock) and the nearest roadblock on that path is closer than d squares (so shorting due to blockage would have been possible for other walkers), a unusual type of walk often results that appears to be possible for both pavilion entertainers and architects (and possibly other roamers): the "confused walk".
   Several beautiful examples of confused walks by an architect were posted by Max as reply 4 to “Predicting Roaming Walks”.  The column labeled “walk-4” in Max's table contains a mixture of distance-aborted walks (when rbc is at 44 or 60), only-blockage-shorted walks (when rbc is at 1 or 15), and confused walks (rbc’s from 16 to 43).  The confused walks in Max’s table begin like blockage-shorted walks, but on their d-th square, they seem to switch into a new walk mode (for 2 to 46 squares of travel) for a strange middle-phase of their walk: neither outward- nor (necessarily) homeward-bound.  In many of Max’s examples, the middle phase of the architect’s walk carries him off the roads.  In others, the architect appears to meekly return part of the way (by road) toward his office before vanishing, but he may just be heading NW and the location of the office is coincidental.  I have been able to reproduce the confused walks that Max reported using pavilion entertainers instead of architects (allowing for minor adjustments need to accommodate the difference between their endurance classes).  In addition, I have been able to create off-road confused walks by teleporters from pavilions provided that their shortest circuits meet the same requirements as for traditional start-square users:  distance to target > d; distance to closest intervening roadblock < d.   As Max's examples illustrate, pavilion entertainers and architects always seem to vanish at the end of confused walks, rather than returning visibly to the walk-finish squre.
   Confused runs unquestionably present a challenging and intriguing subject for research. So far, I have pinned down two new properties and confirmed one predictable property of confused walks.  First, as we would expect, a new walk mode (“confused mode?”) permitting (but not requiring) off-road travel begins on the d-th square of a walk containing such a run.  Second, the total length of a confused walk (not just the confused mode portion) equals the shortest road-connected distance from the roamer’s walk-start square (traditional or teleported) to the roamer’s walk target.  Third, if a roamer in confused mode encounters a building or statue (but not a garden, which he will blast straight through) in his path, he will vanish on the square before the obstruction, rather than detouring around it.
   The third observation has fortunate implications for housing block design.  Since most housing blocks (by the time they are fully developed) are pretty densely packed with structures, entertainers and architects who switch to confused mode in a housing block usually don’t get very far off road before they run into an obstacle and vanish.  Once a confused walker vanishes, he returns promptly to his home building for a coffee break and soon becomes ready for the next walk in his quadramble.  An architect, for example, will not deliberately seek out the one entrance to a block as an escape route so he can indulge in a confused-mode-inspired desire to wander around in the weeds for 50 squares before he is ready to return to the serious business of keeping our temples up.  Less fortunately, an architect patrolling some mines way out in the boondocks is much more likely to neglect his duties while meandering along on a confused walk.
   I confess to being both fascinated and intimidated by confused walks.  Clearly, new analytical tools will be needed to dissect their underlying logic, since classical target squares can only be found on roads, and many interesting turning and vanishing events occur during confused walks at off-road locations that cannot be classical targets.  If I am ever able to make any sense of confused walks, I will report back to this forum in a separate post devoted to them.

Disconnected walk targets.  The default walks of ordinary walkers from traditional walk-start squares look the same whether the walk was defaulted by road absence or by disconnection.  When entertainers teleport from a pavilion with a defaulted walk in its quadramble, however, their behavior reveals the cause of defaulting unambiguously.  Observations of teleporting entertainers make it possible to be more precise (than I dared to be in Ambulomancy) in describing how disconnected road squares default out a quadramble leg.  As the roaming-walker algorithm searches (in the order of priorities shown in Table 1 in Ambulomancy with the correction provided in the Erratum) the region around a shorting center it usually discovers a road square.  The first road square found during the search becomes a provisional walk target.  The algorithm then checks for a road connection to the associated building’s walk-start square.  If a connection is found, the walk target’s status is changed from provisional to confirmed and the target assignment portion of the algorithm ends.  If no road connection is found to the provisional target (i.e, if the provisional target is "disconnected"), the algorithm continues searching the remainder of the search ring (all the squares at the same distance from the routing center as the provisional target) “in hopes” of finding a connected road.  As additional road squares are found, they are each checked for a road connection to the walk-start square.  As soon as a connected road square (in the same search ring) is found, it becomes the confirmed walk target and target assignment ends.  If the search of all the squares in the same ring as the provisional (disconnected) target is completed without discovering any connected road squares, then the status of the provisional walk target is changed to confirmed and targeting ends.
   The greater detail offered by this new description of the targeting process (over the original version in Ambulomancy) can specify the location of a disconnected road square as the target for a defaulted walk.  Although any walker that the algorithm tries to aim at that disconnected target from the usual walk start square cannot get there, a teleporting entertainer can use that disconnected target as a landing square.

Walks defaulted by disconnection.  When the roaming-walker algorithm cannot trace a road connection from the walk-start square to the target, the corresponding walk in a building’s quadramble defaults out.  After appearing on the walk-start square (which may be reached by teleportation) the walker executes a random-mode walk of default length (35 squares for medium-walking entertainers) with the usual default initial directions that apply to default walks beginning on straight roads or corners.  Before posting this discussion, default walks by entertainers from teleported walk-start squares on both kinds of straight roads and all four possible corners were checked to verify that the default directions specified in Ambulomancy still applied.  They do.
   At the end of a default walk, the algorithm sends the entertainer on a (visible) destination walk back to the walk-finish square of the Pavilion by the shortest road-connected route.  If no road-connected path exists, then the entertainer vanishes and immediately is treated as being on a coffee break.  Naturally, entertainers who have teleported onto disconnected walk-start squares are going to have a tough time tracing a road connection back to the pavilion.

Walks defaulted by road absence.   When the usual procedures of the algorithm find a road-square to use as a walk-start point but cannot find a road in the search zone around the relevant routing center to use as a walk target, the corresponding walk of the pavilion (or any other building) defaults out due to road absence.  However, the algorithm also might be unable to find a legitimate walk-start square, i.e., one with a road on it.  This situation only occurs for teleporters, since the pavilion must touch a road suitable for use as a traditional walk-start square or the pavilion’s labor recruiter could not emerge to find the workers needed to operate the pavilion.  Teleporters are different.  They use the walk-target of the previous walk in the pavilion’s quadramble as their walk-start squares, and a walk target will not exist for an absence-defaulted prior walk.  Clearly, the algorithm needs a way to handle a situation in which an entertainer is ready to teleport from a pavilion, but a walk-start square cannot be found for him/her.
   When the roaming walker algorithm cannot use the walk-target of the previous walk as the walk-start square for a teleported walk, the algorithm puts the affected entertainer on an “emergency walk-start square”: the #1-priority search square around the routing center for the walk before the previous walk, and tries to use that square as a walk-start point for the entertainer.  Thus, a musician who is scheduled for a teleportation-begun walk to the NE from a pavilion with a absence-defaulted NW walk (i.e., with no walk target for its NW walks, and, therefore, no standard teleportation walk-start square for its NE walks), will appear on the emergency walk-start square, one square due north from the SW routing center.  If a road exists on the emergency walk-start square, the walker can execute any kind of walk (grounded, defaulted, shorted) from it under the usual rules applicable to the local road geometry.  However, no road square is required on the emergency walk-start square for an entertainer to appear there.  If an entertainer appears on the emergency walk-start square and that square contains no road, the entertainer vanishes after a very few animation frames of time have passed.  Whether the dancer’s emergency walk start contains or lacks a road, and whether she, therefore, makes a walk or disappears in a animation frame or two, her appearance on the emergency square uses a walk from the pavilion’s queue and advances the formal direction for the next walk by 90 degrees clockwise.
   If the emergency walk-start square of a dancer in a three-man jump team contains a road, that square will simultaneously serve as the ordinary teleporting walk-start square of the musician, since that is the highest priority search square around the the previous walk from the musician's perspective.  The algorithm starts both entertainers on that square at the same apparent time for their walks.  The two walks are likely to be somewhat different, however.  The musician will be on a walk that has been defaulted due to road absence.  She may, however, be able to trace a connection to the walk-finish square of the pavilion, in which case she will visibly walk there (in destination mode) at the end of her 35-square random mode default walk.  The dancer may or may not have a walk target around the pavilion’s NE routing center, and that shortest (road-connected) route to that target may or may not be roadblocked.  The dancer can, therefore, execute nearly any conceivable walk from the emergency walk-start square - provided that it contains a road.

   Teleportation from a pavilion with a walk defaulted due to road absence can be difficult to visualize from a description, so Fig. 2 shows an example.  Fig. 2 illustrates the critical squares that apply to the walks to be executed by the members of a three-man jump team who are about to end their coffee break in the pavilion shown.  Acting as the spotter, the juggler appears on the usual walk-start square of the pavilion (yellow road in Fig. 2) to begin a walk with a formal direction (in this example) of southwest.  The musician begins her NW walk by appearing on her primary teleporting walk-start square (yellow plaza), which happens to be the same square that the juggler is using as the target of his SW walk.  The musician has no target for her NW walk, since no road squares appear in the pavilion’s NW routing zone (desert squares), so she will be making a default walk.  The dancer would normally use the musician’s NW walk target as her primary teleporting walk-start square, but the musician has defaulted due to a lack of a NW target.  Consequently, the dancer resorts to the emergency walk-start square for a walk with a formal direction of NE which falls on the #1 search priority square near the SW routing center and just happens to coincide with the same yellow plaza tile that musician uses as her primary teleporting walk start.  The musician and dancer appear on the yellow plaza tile simultaneously and the game superimposes their images.

Figure 2.  Selected critical squares for a three-man jump team leaving a pavilion with an absence-defaulted northwest walk.  The spotter (the juggler) starts a southeast walk.  Pop-ups on colored road and plaza tiles specify walk-start, -target, and -finish locations.  Small statues mark routing centers.  Much of the road-free NW routing zone is highlighted in desert.

   The dancer and the musician do not execute the same walks.  The musician (on a default walk) leaves the yellow plaza tile heading towards the NE (the default direction for a west corner).  Meanwhile, the dancer, who uses the green plaza at the top of Fig. 2 as a walk target, shorts out due to the presence of the roadblock on the shortest route to that target.  On this blockage-shorted walk, the dancer follows the route to her target as far possible until the roadblock prevents further progress, and then completes the rest of her (total length = 35 squares) shorted walk trapped in the little elbow of road to the left of the roadblock wandering aimlessly in random mode.  Thus, the dancer moves SE on her first square of travel, so a difference between the walks of the musician and dancer reveals itself immediately.  The juggler’s SW walk also shorts out due to blockage, because the same roadblock that stopped the dancer also gets between his walk-start (yellow road) and his target (yellow plaza).  The juggler, therefore, proceeds in destination mode straight along the route to the yellow plaza square as far as he can without hitting the roadblock, and then converts to random mode (without reversing course) to use up the remainder of his 35-square shorted walk.
    At the end of their blockage-shorted walks, the juggler and dancer both vanish and return to the pavilion immediately to start a coffee break.  The musician does not vanish, because she is on a default walk and has a road-connected path to the pavilion’s walk-finish square.  At the end of random mode, she converts to destination mode to get to her walk finish, which lets her pass through the roadblock.

Certainly, more research remains needed into the behavior of pavilions and their walkers, especially into confused walks from these structures.  Other walkers in the game also have been documented or reported to teleport, notably bazaar sellers, who seem to require a kind of a “collision” to initiate teleportation, which sounds remarkably similar to the initiation of teleportation from pavilions and is a subject crying out for a bit of investigation.   Although they have not been mentioned in this post, Max’s table showing confused walks (reply 4 in "Predicting roaming walks") also provides examples of “false starts”, doubling back by a roamer leaving a walk start in a dead end after a single square of travel.  Given the right roads, entertainers leaving pavilions may also give false starts.  Teleporters landing on at least some dead ends probably exhibit false starts, too.  If I ever figure out how both false starts and confused walks work, then I think we will be ready for a second edition of Ambulomancy and another look at pavilions.  Until then, I hope this post gives readers a better idea of how teleportation - at least by dancers and musicians - works.

StephAmon

[This message has been edited by StephAmon (edited 08-23-2005 @ 05:12 PM).]

StephAmon,

This is great stuff! I have a few comments. (I thought it would be good to collect data on unusual walker behavior here, for lack of a better place.)

For those of us who are really lazy, reply #1 should include a link to your Ambulomancy thread, Predicting Roaming Walks.

You imply that tax collectors don't wander off into the wilderness like architects, but I've noticed tax collectors doing that fairly often. (Since tax collectors are typically found in crowded housing blocks, they often encounter a solid object and disappear immediately upon turning away from the "normal" route.)

Entertainers (at least musicians) can teleport when leaving their "school". This has caused me problems only once or twice, and it was easy to correct (by moving the conservatory a little), but it would be nice to know when and how it occurs.

When both traders from an upgraded bazaar are generated at exactly the same time, one typically teleports (while the other appears in the normal place). From my limited observations, I'd guess that there are two places that bazaar traders may teleport--one is to a square fairly near the bazaar (perhaps a previous walk target, like for entertainers), and the other is to a previous destination (storage yard or perhaps granary) of one of the bazaar's buyers. Also, I've seen a report of one bazaar trader generated (in the normal place) when two (one of them teleporting) would have been expected (which might mean that the teleporting trader appeared on a roadless "emergency walk-start square" and quickly disappeared).

"Roaming" entertainers in general (including senet players and jugglers from booths) sometimes disappear instead of returning to their buildings. They may display other unusual behavior, but I haven't noticed any.

[This message has been edited by Brugle (edited 08-23-2005 @ 07:24 PM).]

Bravo!

A fantastic piece of work, StephAmon. I am deeply impressed.

Jimhotep:

Thank you for noticing and commenting. I also owe you a bunch of thanks for rekindling my interest in roamers. Posting Ambulomancy was a big enough effort that I was pretty well burnt out until you started raising questions that I could almost answer.

Brugle:

I like the idea of collecting unusual walker data in a single convenient spot for easy reference. Maybe the information would make a good table: something like a “Table of Bad Habits”. Each row could correspond to a different walker and we could assign each bizarre habit to a different column. Different letters as table entries could indicate things like whether a walker just vanishes or ghosts home.

I think I might be able to tackle such a task without a lot of trouble right now, because I have *.sav files of lab setups with convenient roads for checking many of the properties - as long as I do not wait so long that I forget the all the file names. I am kind of on the verge of burning out again right now, not so much from fooling around with teleportation, but my first nine hours of trying to figure out confused walks were a nightmare. Arrgghh! Why don’t the little vermin just stay on the roads and save us all a lot of aggravation?!?

Tax collectors in the weeds? I might have known these guys would cause more trouble. I’ll check. I also want to check senet players. Somebody on the design team might have thought it was appropriate to have the game’s big drinkers get lost and fall off the roads.

I can independently confirm your observation that jugglers from booths vanish sometimes. I also stuck one in Max’s off-road-inducing geometry and he went off road just like he came from a pavilion. Still need to check folks from bandstands, but I’d be astonished if source venue influenced (non-teleporting) bad habits at this point.

Two weeks ago, I would not have known what you were talking about with your observation of entertainers teleporting from their schools. Now, I do. I saw it with dancers leaving their school. I was tempted to drop what I was doing and try to learn more about it, but (for once) sanity (or self discipline) prevailed, and I managed to persuade myself that these guys were destination walkers and, therefore, not my department. Might dance schools simply use a different priority order for walk-start squares? One that assigns higher priorities to some squares further away from schools than to other squares right next to the schools? Sort of like the non-adjacent walk-finish squares you found for temple complexes? I am not sure I am the right guy to tackle this project. The last time I tried to figure out walk-start squares (for bandstands), I screwed up and Jimhotep had to save me.

I may take a shot at teleportation from bazaars before too long. If they work a lot like pavilions, they should not be too bad. If the conditions needed to initiate teleportation are different, however, then they might chew up a prohibitive amount of research time. We’ll see.

I’ll come back here and post again if I learn more about any of these phenomena.

For now, thanks to both of you for reading my (way-too-long) post.

StephAmon

StephAmon,

My Menat Khufu has a tax collector who'll go "in the weeds". In October, the tax collector in the northeast block will leave the loop road (through a roadblock) and disappear fairly soon (when he tries to go through a house), and on his next walk he'll simply disappear (when he tries to go through a temple). If the buildings in his way are deleted beforehand, he'll go a few tiles over bare ground, disappear, and invisibly return to his building (without registering houses for taxation). When returning to his building invisibly, he can be tracked as a "red spot" in the Administration/Tax income overlay, and he appears to take the shortest route back to his building, travelling diagonally if necessary (like your "ghosting architects").

Quoted from reply#8:

Sorry, that excuse won't do , since you've already dabbled in the subject:

Quoted from reply#1:

I suppose "teleporting" may not be the best term to describe an entertainer who appears from its "school" at an unusual place. As far as I can tell, when this occurs at a "school", it occurs for all entertainers generated by that "school". Teleporting traders from upgraded bazaars and teleporting entertainers from pavilions typically occur only part of the time, when specific conditions are met.

Another couple of questions, not related your previous post:

Quoted from reply#4:

If an entertainer appears on the emergency walk-start square and that square contains no road but an adjacent tile contains a road, will she disappear (as described) or walk onto the adjacent road?

Quoted from reply#1:

I'd guess that the condition should be whether or not the arriving entertainer's stage was occupied when the entertainer left its school (not whether the stage is occupied when it arrives at the pavilion). The same guess applies to entertainers arriving at a bandstand (with the north tile of the music stage replacing the pavilion).

I hope you don't get too burned out. It's happened to me, as forumers who've followed some of my walkthroughs (and waited months for the next installment) know.

[This message has been edited by Brugle (edited 08-24-2005 @ 02:38 PM).]

I am struck by how walk-targets are chained when teleportation is present; one walker's destination point is the next walker's starting point. (Modify the story appropriately for "backup" target squares in case someone defaulted due to absence). I wonder if this feature could be part of the explanation for why entertainers sometimes appear a distance away from their schools. Are the schools near a target square of the associated venue? If so, then it seems that the walk-start square of the entertainer has somehow gotten linked up with a target square from the venue. I wonder if a bazaar trader will appear near a SY if that was the previous trader's destination?

The "chain of walk-targets" seems to be an underlying feature in how walkers normally behave; frex a grounded walk starts a random walk on the target square of its destination walk. Walkers typically become destination walkers at their "go home" part of their journey. The same mechanism, in a distorted form, seems to be present in teleportation.

StephAmon has noted that if a bandstand is destroyed and rebuilt in a different configuration while a musician is en route, it will subsequently produce two musicians. Both of these facts are remarkable; that a musician will serve in a different bandstand than the one it was trained for, and that a bandstand will produce multiple musicians.

How far away can it be built, I wonder? I bet the musician has to end up somewhere in the venue; but does it necessarily have to be adjacent to the north tile of the music stage?

That reminds me of something.

I want to report a tidbit that is already probably well known, but that I don't recall seeing written down anywhere. It is well documented that the interior roads of a pavilion can be disconnected from the rest of the road system; entertainers and labor seekers will enter and exit through the sides of the dance stage. (Note that the normal priority for a walk-start square are suspended; if the NE side of the dance stage is in the pavilion, and the SW side is adjacent to a road, the labor seeker will leave by the SW side.) Well, the same thing is true for a bandstand. If the northern tile of the music stage is adjacent to a road, labor seekers will exit there, and entertainers will enter there, in manner analagous for a pavilion.

Speaking of those pavilions, I built one in the usual way and built a dance school nearby. Then I built another dance school that was on a disconnected road that touched the dance stage. No dancers were dispatched. Only when the road systems were connected did a dancer appear. It entered by the dance stage, however, and didn't go around to where the roadways joined. Destroying and rebuilding the pavilion eliminated this behavior. I didn't try all the permutations, but it did strike me as odd. It also offers another hint at the connection between schools and venues.

Those dancers remind me of something else. Something I was going to watch for, but never got around to doing. StephAmon notes that when there is no show going on, an entertainer enters by the walk-start square. If there is a show going on, it enters by the walk-end square. I wondered if that was also the same with buildings that are animated when they receive supplies; potters and weavers and such. But I never got around to checking.

So many interesting things occur to me when my Pharaoh disk is on the other computer! I hope I didn't get the details wrong. Was it a pavilion or a bandstand that I tried to serve with two schools?

Brugle:

Ouch! You've got me dead to rights on destination walkers. But, I still think I'd be likely to make a hash of it if I tried to figure out teleportation from schools.

I think you are also right about the need for conditions to be met to enable teleportation from bazaars and pavilions, but I'm not sure about dance schools. I saw every single dancer leaving her school "teleport" onto the same square of the same road, as long as that road was available (a few dozen dancers).

I do not know if a musician or dancer will move a single square to get to a road from a road-free emergency walk-start square. It sounds like something that would be worth checking. I do know that she will not move two squares to get onto the nearest road.

I would wager a big stack of debens that you are right about the time at which the destination-walking entertainer is assigned her arrival square for a pavilion or bandstand. If a second juggler left his school on the way to a pavilion before the first juggler from the same school got there, the destination-walk of the second juggler would be generated (and perhaps "locked in") while the juggle platform was unoccupied. Interesting and checkable. I would not have seen this effect because all my schools were quite close to the pavilions I played with. (So close, in fact, that arrriving dancers teleported directly from school to pavilion at one point!)

As much as I love walker research, I think I had better take a break and actually play the game for a bit. You would not, by any chance, know how I could get my hands on a copy of Djed Djedi's spreadsheet would you? The links in his "New City Designing Tool" thread no longer seem to work. I would love to eliminate the need for entering maps by hand into PDC if I could.

Many thanks for the great ideas.

Hi StephAmon!

I've just forwarded a copy of Djed Djedi's spreadsheet that I got from Vitruvious.

Have fun playing!

(I'm going to try WinterPharaoh's block sometime soon...)

Thank you, thank you, thank you!

I can tell there's not going to be any research in the walker labs tonight!

I have been messing about in Heh, but the whole time I was vexed by the incomplete status of our knowledge about which walkers are confusible, vanishing, or otherwise possessed of bad habits.  After all, we were so close to being done.  So, I went back to the lab and got some more data.  I have summarized the information I have available (plus a bit more that I recycled from earlier work or stole from Brugle's "Random walker start and finish points" for completeness) in the table below.

The table shows one column to indicated whether a particular kind of roamer makes confused walks, but it includes no separate columns to show whether the walkers make walks aborted by distance and walks shorted only by blockage.  This is because in my observations, these three kinds of walks seem to be a kind of "package deal"; if you get one, you get them all.  All three of these walks are terminated by vanishing.

I am pleased to offer independent corroboration of Brugle's observation of ghosting by tax collectors.  The latest results indicate then whenever a tax collector or architect vanishes, he ghosts home as described in earlier posts to this thread.  In contrast, I have not been able to detect ghosting by any of the entertainers (and I have checked them all) even though they execute the same three unusual kinds of walks (including confused) that architects and tax collectors do.
 

Table.  Habits of roaming walkers.
 

Walker	Recruits labor	Confusiblea	Ghosts after vanishing	Default walk length	Walk-start typeb
Fireman	Yes	No	N.A.	43	Common
Policeman	No	No	N.A.	43	Government
Magistrate	No	No	N.A.	43	Government
Tax collector	Yes	Yes	Yes	35	Government
Architect	Yes	Yes	Yes	43	Common
All entertainersc	Yes	Yes	No	35	Commond
All otherse	Yes	No	N.A.	26	Common or specialf

^aConfusible walkers also execute and vanish after walks aborted due to distance and walks shorted only by blockage.
^bSee Brugle's "Random walker start and finish points" for differences between common and government walkers.
^cMusicians, and dancers execute confused, aborted, and solely distance-shorted walks from either traditional or teleported walk-start squares.
^dAll entertainers leaving bandstands treat that venue as a 1x1 building located on the northern square of the music stage; all entertainers leaving pavilions locate their walk-start squares relative to the 2x2 dance stage.
^eProperties of plague walkers are still unknown.
^fSee Brugle's "Random walker start and finish points" for labor recruiters from storage yards and walkers from temple complexes.

I will include the above table in the next edition of Ambulomancy, which may be coming out sooner than I thought.  I think I have caught myself in another error, and I would hate to have to post another erratum.  Until then, it's back to Heh, where I think I may be able to create a plague carrier right where I want him.

StephAmon

StephAmon,
It appears that plague has a chance of occuring randomly at the start of a month when city health is bad enough, and when it occurs it infects the house with the worst disease risk at that moment. The plague walker appears in a road tile adjacent to the house (or 1 tile away from the house, if there are no adjacent road tiles), probably using a common rule (such as the walk-finish point for most random walkers).

Brugle:

Thank you for the tips. They helped. In fact, I would not have had the patience to keep backing up, tweaking the roads, and trying again to get more plague carriers without your assurance that there was a random component to the carrier's appearance. In fact, my observations supported every one of your rules regarding plague. I have also recorded enough new observations that I thought it might be helpful to put them in a separate thread with "plague" in the title for the convenience of future players who - perhaps in an emergency - want to learn a bit more about plague carriers.

I'm just ironing out the last couple of details right now. It sure would be nice if I could find a firm rule governing the initial direction taken by a plague carrier if he appears in a straight road or on a corner similar to the rules for default walks by ordinary roaming walkers in the same road geometries. I already know, however, that if such standard initial carrier directions exist, they are not the same as default initial directions.

Confused walks, as described by StephAmon in reply #3, occur when certain walkers (architects, tax collectors, and entertainers) have the initial phase of their random walk shorted due to both distance and blockage. The confused part of the walk begins after the walker has gone the default walk length. During the confused part of the walk, the walker may travel off-road, but it disappears if it tries to go through a building.

I can explain all confused (including off-road) walks (by architects, tax collectors, and senet players) that I have examined. The explanation is a fairly simple combination of only-distance-shorted and only-blockage-shorted walks by confusable walkers and seems right to me. I did not examine confused walks by teleported entertainers (or any non-senet entertainers).

A confused walk begins the same as any blockage-shorted walk: by following the shortest route to the walk-target until a roadblock is encountered, then "roaming" until the default walk length has been traveled. At this point, instead of returning to its building (like a non-confusable walker) or disappearing (like a only-blockage-shorted confusable walker), the walker follows the directions for the remainder (after the default walk length) of the shortest route from the starting square to the walk-target square. During the confused part, the walker may go through rocks but not through buildings or over water. (I did not check whether entertainers are blocked by water but not rocks, but I assume that their behavior is the same as architects and tax collectors.) The walker disappears after the confused walk (if it hasn't already).

Consider this example (with north up-left):


With no other roads nearby, the walk-target tile for the tax collector on his SW walk is the end of the dead-end road. The route to the walk-target is longer than the tax collector's default walk length (35 tiles) and a roadblock is encountered before going the default walk length, so he will take a confused walk. He takes the route to the walk-target up to the tile next to the roadblock, then roams around the loop until he has gone the default walk length, which puts him on the road tile next to the garden (where the confused part of the walk begins). If he had taken the route to the walk-target, he would have reached the plaza tile after going the default walk length, and so he follows the directions for the remainder of the route to the walk-target: NW for 6 tiles then SW for 2 tiles. He goes NW (through the garden) for 6 tiles and SW for 2 tiles (if nothing blocks his way) and then disappears.

The behavior of confused walkers suggests that routes are stored by the game as absolute directions (such as go NW, go NE) rather than relative directions (such as go straight, turn right).

Interesting, and once again I find myself wondering how on earth you came to make such an unobvious observation

I can verify that destination walker paths are indeed stored in the manner you describe - a simple array up to 500 bytes long containing a sequence of numbers 0 through 7 (representing the directions NE through N clockwise, including semicardinals). I mentioned this in reply 14 of this thread in the C3 Scenario Designers' Forum

It's also true that, once computed, the entire path is stored even though the walker may have completed part of it. What I mean by that is that when a save is made and reloaded the whole path is there, and the walker finds his current position along it by referring to the tile-counter which is stored as one of the fields in his walker Table entry.

It would appear that this class of walker, despite having been deflected from his 'destination' walk and travelled some distance in 'random' mode, picks up his destination path again at his current tile position. Since that path is no longer relevant, I would think it fair to call this behaviour unintended (i.e. a bug).

This class of walker is one who would in fact have reached his walk target despite it being beyond his normal 'default' walk length had it not been for the road-block. I think therefore that StephAmon's assertion that the walk is distance-shorted is semantically incorrect. It is shorted only by blockage, but there appears to be a coding conflict where an 'end of default walk' reaction is signalled when his tile-counter reaches 35 while part of his 'state' believes he his 35 tiles along a destination path. This behaviour is only observed on (what would have been) 'long' walks - on 'normal' blockage shorted walks he turns to face his home building, disappears and returns invisibly.

One does wonder why this class of walker has these idiosynchracies - the same class spawns 'invisibly' on its building's normal starting square then walks diectly to its destination square to start its 'physical' walk.

Perhaps these walkers were never intended to do the 'long' walk in the first place, and perhaps they didn't 'fix' it after C3 because they were happy with the effect.

Quoted from Trium:

how on earth you came to make such an unobvious observation

That's funny. I thought it was quite obvious.

I was on an airplane, trying (but failing) to get a little sleep, and decided to think about confused walks since I had read StephAmon's description recently (and used it in designing my new Iunet). When a blockage-shorted confusable walker has finished the roaming part of the walk (having gone the default walk distance), it appears confused (wandering around in a non-trivial way, often off-road) for the number of tiles (if any) that would remain in its walk if it had simply walked to the walk-target. It seemed to me that the simplest way of implementing such confusion would be to follow the directions for the remainder of the walk to the walk target. When I realized that this would make distance-shorted confusable walks the same except for some roaming in the middle if a roadblock is encountered before the default walk length has been travelled, I was pretty sure that it was correct. (I had to wait over a week before testing it. )

But I had one reason for doubt: if my explanation was true, it seemed likely that StephAmon would have tried it out and reported it. I guess it isn't as obvious as I thought.

Quoted from Trium:

I mentioned this in reply 14 of this thread in the C3 Scenario Designers' Forum

I missed your latest posts in the "Saved games tool (C3GameExplorer)" thread. Thanks for calling it to my attention.

Regarding a limit to the number of destination walkers: I recall someone (I think from Impressions) saying that there were 3 "size" limits: buildings, walkers, and point-to-point routes (or something like that).

Quoted from Trium:

I would think it fair to call this behaviour unintended (i.e. a bug) ... Perhaps these walkers were never intended to do the 'long' walk in the first place, and perhaps they didn't 'fix' it after C3 because they were happy with the effect.

Perhaps. On the other hand, there are a lot of subtle differences in walker behavior. I wouldn't be surprised if the designers knew what they were doing.

Consider the walkers that are confusable. Architects will cover a lot of buildings before wandering around in the weeds, and a collapse is not nearly as bad as a fire. The worst that can happen with a confused tax collector is a little tax is uncollected. Entertainers are important, but aside from senet players and zoo keepers (who don't exist in C3, where this behavior began), entertainment can also be supplied by walkers going from a "school" to a venue. The selection of these walkers as confusable is at least a little evidence that the confusion was deliberate.

Quoted from Trium:

This class of walker is one who would in fact have reached his walk target despite it being beyond his normal 'default' walk length had it not been for the road-block. I think therefore that StephAmon's assertion that the walk is distance-shorted is semantically incorrect.

StephAmon also calls a confusable walker who reaches his walk target after a walk that is longer than the default walk length distance-shorted, so I guess you'd say that that is also semantically incorrect. Perhaps so. This passage from Ambulomancy suggests that StephAmon considers shorting to be anything (other than non-connection) that prevents a grounded walk (walking to the walk-target then roaming for the default walk length):

Quoted from StephAmon:

When a blockage-free walk-target has been identified, the algorithm compares the length of the shortest circuit to d. If shortest-circuit length is strictly less than d (not “less than or equal” to d), then the target is deemed to be walkable, no shorting occurs, and a grounded walk (described below) is issued.

(I did not search StephAmon's old posts for a precise definition.)

"Shorting" is probably not the best term for whatever StephAmon means when he uses it, but it's not too bad.

Quoted from Trium:

there appears to be a coding conflict

Only if you assume that the code programs unintended behavior.

Quoted from Trium:

This behaviour is only observed on (what would have been) 'long' walks - on 'normal' blockage shorted walks he turns to face his home building, disappears and returns invisibly.

I'm pretty sure that I saw a tax collector turn to face his building, disappear, and return invisibly after a confused walk. The only difference from a 'normal' blockage-shorted walk is that the 'normal' walk has already gone the start-to-walk-target distance after roaming but the confused walk has not.

Quoted from Trium:

One does wonder why this class of walker has these idiosynchracies - the same class spawns 'invisibly' on its building's normal starting square then walks diectly to its destination square to start its 'physical' walk.

Not all of them do that. A tax collector starts his 'physical' walk at his destination square (as does a C3 engineer), but an architect might not and a senet player might not.

[This message has been edited by Brugle (edited 09-01-2009 @ 12:41 PM).]

This is an amazing piece of research. I'll have to read it several times more to fully understand it.

Now all we need are "gamey" explanations for the above (like dancers don't like to be seen hanging out with jugglers, and sneak out the back if they happen to leave their stage at the same time....)

Henipatra

A small correction. From reply #1:

Quoted from StephAmon:

If the stage for his/her type of arriving walker is unoccupied, then the walker disappears on the walk-start square of the dance stage.

The walker will disappear on a road tile (that may be roadblocked) which is adjacent to a side (not just a corner) of the dance stage and is the first such road tile encountered when going clockwise around the dance stage from the tile N of the N tile of the dance stage. That tile will be the walk-start square of the dance stage only if it is not roadblocked. For details, see Entertainer going from school to venue start and finish points.

Quoted from Stephamon:

If no road connection is found to the provisional target (i.e, if the provisional target is "disconnected"), the algorithm continues searching the remainder of the search ring ..... As soon as a connected road square (in the same search ring) is found, it becomes the confirmed walk target .... If the search of all the squares in the same ring as the provisional (disconnected) target is completed without discovering any connected road squares, then the status of the provisional walk target is changed to confirmed and targeting ends.

I've always doubted this. I don't have Pharaoh on this machine, but I recall observing (and can confirm in C3) that the first tile found determines the walk target. If it is disconnected, a default walk ensues regardless of whether any connected tiles are in the same 'search ring'. I use this regularly to prevent walkers taking routes I don't want them to.

Perhaps I mis-remember, but I don't think Pharaoh is any different

[This message has been edited by Trium (edited 01-26-2010 @ 11:26 PM).]

Quoted from Trium:

I don't think Pharaoh is any different

In this respect, Pharaoh and C3 are different. StephAmon is correct.

When I advised some C3 players to read Ambulomancy, saying something like "C3 random walkers have similar behavior", I wondered if I should mention that difference but decided against doing so.

In a large housing block, I've tried making a pavilion one square away from the loop and putting a roadblock in-between; putting a single house next to the roadblock for access to workers. It still gives a large desirability boost to surroundings and entertainment points but doesn't give the headache of messing up housing blocks, at least on normal difficulty.

Building my Immortal Iunet, I saw a type of confused walk that I hadn't read about.

Consider a walker that has a route from its start square to its target square but can go no more than 1 tile without being blocked from any wandering. (A walker could start in a non-roadblocked or roadblocked square and not be able to wander anywhere, or it could start in a roadblocked square, walk into an adjacent non-roadblocked square, and not be able to wander anywhere.) A non-confusable walker will start walking to its finish square at that point. However, an architect or tax collector or entertainer will become confused at that point. It doesn't matter whether the route to the target square is shorter than the default walk length.

If the walker becomes confused at the start square, or if the walker walks 1 tile along the route to the target square and then becomes confused, then the walker will follow the route to the target tile and vanish. To a casual observer, the walk will seem fairly normal (except for going through roadblocks and perhaps going a long distance). In my Immortal Iunet, one of the architects makes walks along roads (including roadblocks) of 9, 15, 71, and 165 tiles, in each walk becoming confused after 1 tile.

There is a possibly-related fact suggested by StephAmon in reply #3 (or the second edition of Ambulomancy) together with reply #14. If the distance to the target square is greater than the default walk length, and if there are no roadblocks along the route before the default walk length has been traveled, then an architect or tax collector or entertainer will walk to the target square and vanish. (Caesar III players have used the similar behavior of engineers to make them go a long way.)

Quoted from Brugle:

There is a possibly-related fact suggested by StephAmon in reply #3 (or the second edition of Ambulomancy) together with reply #14. If the distance to the target square is greater than the default walk length, and if there are no roadblocks along the route before the default walk length has been traveled, then an architect or tax collector or entertainer will walk to the target square and vanish.

I quoted the evidence but didn't make the obvious conclusion. (By the way, there is a slight error above--"greater than" should be "greater than or equal to".)

The long walk described above (that architects, tax collectors, and entertainers can take) is actually a "confused" walk. (It isn't obviously confused because it follows the roads.)

As far as I know, this is how confusion occurs. An architect, tax collector, or entertainer begins to walk to its walk target. If it encounters a roadblock, it shifts to roaming mode. Whether or not it is in roaming mode, if it goes the default distance before or at the same time that it reaches its walk target, or if it cannot go further (because it is hemmed in by roadblocks), then it becomes confused. It follows the rest of the directions (if any) to its walk target, as described in reply #17, and then disappears.

So simple. Why didn't I see it before?

Quoted from Brugle:

If the distance to the target square is greater than the default walk length, and if there are no roadblocks along the route before the default walk length has been traveled, then an architect or tax collector or entertainer will walk to the target square and vanish.

This what I was looking for, and I think I can make senet players, zoo keepers, architects and tax collectors take long trips everytime. I made a test with the zookeeper in my Hetep24k and it worked.

The rule of thumb: If we can make the road between the walk start to the target square (there are 4 target squares, cycled clockwise every spawn) of these "can become confused" walkers free from a roadblock, give them enough road tiles to walk, so that they can finish their given walk length (26, 35 or 43) before the roadblock.

Let me write my version of ambumolancy.

* Each walker is given such amount of "energy" (or pocketmoney or whatever), that will be spent or used 1 if he/she moves from a road tile to the next. It's symbolized as d (default walk length) in the StepAmmon's Ambumolancy, where d can be 26, 35 or 43.

* From the starting square (usually his/her building's start point) the walker must take the shortest path to 1 of his/her 4 target squares (more about these target squares can be found in StepAmmon's ambumolancy or in this forum, and more about the starting square can be found in earlier post in this thread).

* In this walk the walker can't pass a roadblock.

These 3 facts are same with StepAnnom's version, but calling the default walk's length as an "energy" will bring us to a new perspective. When StepAmmon said 4 cases would occur, here there are only 3:

* Case 1: If a walker reachs the target's square before the "energy" running out, he/she becomes a gift: his/her "energy" is back full and can walk wherever he/she wants (random walk), but can't pass roadblocks, until the energy is empty and returns home in the end (takes the shortest path to building's finishing point and ignores roadbloks).

* Case 2: If the energy runs out before the walker reachs the target square:


* Case 3: The walker can't continue his/her walk to the target square because of a roadblock. Both type of walker starts to walk randomly like case 1 but there is no gift of "energy" for him/her now, and when the "energy" runs out:

Quoted from Brugle:

The long walk described above (that architects, tax collectors, and entertainers can take) is actually a "confused" walk. (It isn't obviously confused because it follows the roads.)

With this new perspective (case 2 for architects, tax collectors and entertainers) I don't call this long walk as a confused walk. I would say it is a braver walk, when the others return home (coward), the architects, tax collectors and entertainers continue to finish the given order: walk to the target square, although they will be dead in the end. The only confused walk is in the case 3: a roadblock makes them confused.

So, what do you think?

Quoted from andib2n:

he/she becomes confused, ... There must be a pattern where he/she walks, but I don't know about it yet

The pattern is described in reply #17.

Quoted from andib2n:

what do you think?

Your explanation seems to me to be more complicated, but if it appeals to you that's fine.