AI comparison chart on the wiki

[Zuu]

I like your idea and the work that have been done so far. Maybe I'm a bit late to the scene, but here is my opinions on how it could be improved:

I would myself opt for removing the table with version info and direct download links. It doesn't contain info on dependencies nor how to install the AIs. So how will the download links be of any help? The links to the AI threads can be moved to the AI names in the feature table and the Other AIs list.

May I suggest to rename "NewGRF Support" into a something that informs the reader that only AIs with lack of NewGRF Support will have a note in this column. Eg. "NewGRF Compatibility" or "Known NewGRF Problems" etc.

[Brumi]

May I suggest to rename "NewGRF Support" into a something that informs the reader that only AIs with lack of NewGRF Support will have a note in this column. Eg. "NewGRF Compatibility" or "Known NewGRF Problems" etc.

That's a valid point, I changed it to 'NewGRF compatibility'.

I would myself opt for removing the table with version info and direct download links. It doesn't contain info on dependencies nor how to install the AIs. So how will the download links be of any help? The links to the AI threads can be moved to the AI names in the feature table and the Other AIs list.

As Alberth has said, version information should be kept as a reference. Of course it may be moved to the features table, like 'SimpleAI (v6)'.
Originally I made two different tables because it is done that way on most Wikipedia comparison articles (like http://en.wikipedia.org/wiki/Comparison ... il_clients as an example) from which I took my inspiration. It's also OK for me if there is only one table. I don't have the time to do it right now, and I won't be available tomorrow, but it's a wiki after all

[Felix Atagong]

As you may have noticed I am currently playing OpenTTD and - en passant - noting bugs and posting them on this forum.

As some AIs haven't been updated for a long period (despite some bugs) it would be nice to give some 'quality' marks at the wiki, something in a range from 1 to 5, for instance. With those AIs that are guaranteed to crash a 1 level and those that are stable a 5.

This could help the unfortunate newbie who is now watching at a forest of AIs.

[planetmaker]

It would be nice if you could mention in the wiki page that a certain AI is no longer updated (or hasn't been updated since...)

I'd argue against adding that kind of information - it will be very quickly outdated old very quickly. And when the current AI version on bananas is as mentioned in that wiki page, it is clear that there hasn't been an update since; and you see the wiki page's history, if you want.
But even if added: what information does it give you? Important is, that the AI works satisfactorily without crashes.

[Felix Atagong]

Sorry, didn't make me quite clear then, that is exactly the information I would like to find on the wiki...
I don't think that a dead AI will suddenly resuscitate... and if it does... so the better...

[planetmaker]

Sorry, didn't make me quite clear then, that is exactly the information I would like to find on the wiki...
I don't think that a dead AI will suddenly resuscitate... and if it does... so the better...

The problem lies in the definition of "dead". Is an AI that works - but has not been updated for 15 months - a dead AI? Why? Why not? What does 'work' mean? Crash? How often and under which circumstances? Is dead only if it crashes and hasn't been updated? Whatabout a randome very seldom-triggered crash on a otherwise still very competitive AI? etc pp.

[Brumi]

It would be nice if you could mention in the wiki page that a certain AI is no longer updated (or hasn't been updated since...)

Basically the last release date tells you that. It can give you an idea how 'active' an AI is, but it can be misleading, because I can imagine a super-stable AI that doesn't even need to get updated.

Currently you can measure stability and activity by simply looking at the forum threads. Those that have a lot of bug reports since the last update without the author replying to it can be considered inactive. But I think we cannot expect new users, who have just downloaded the game and would like to get some AIs to play against, to read several pages of discussion on the forums.

I think if we would like to measure the stability of AIs, it cannot be done in some scientific way, rather in an empirical way, based on users' experiences. Some stability measurement would be useful in my opinion, but whose responsibility would it be to do that?

[planetmaker]

I think if we would like to measure the stability of AIs, it cannot be done in some scientific way, rather in an empirical way, based on users' experiences. Some stability measurement would be useful in my opinion, but whose responsibility would it be to do that?

I think you have a very important point there. I summarized my thoughts on that matter in this topic: http://www.tt-forums.net/viewtopic.php?f=65&t=55915 as it might go too far here and definitely is worth a discussion on itself.

[Hephi]

for http://wiki.openttd.org/Comparison_of_AIs

Can someone post the 'rules' of what needs to be working to be considered save/load compliant? Thanks
Like my AI was scored NO but when it loads it makes it's list of routes and vehicles from data gotten from the game. Only thing it doesn't do is completing unfinished routes if the game was saved after construction started but before the route was inaugorated (when road/railfinder is running).

[Zuu]

for http://wiki.openttd.org/Comparison_of_AIs

Can someone post the 'rules' of what needs to be working to be considered save/load compliant? Thanks

Wiki article on save/load: http://wiki.openttd.org/AI:Save/Load (it is not a rules-list, but help on how to use the save/load-system)

In my AIs I just store an empty table in the Save()-function so that OpenTTD will not complain about that the AI has not implemented saving. I do like you, reading all connections from the map instead of relying on saving data in the save file. This has the additional benefit that you can save a game after the AI has crashed, load it and recover from there. (though, one should still fix the bug, it can sometimes help in debugging/testing)

Only thing it doesn't do is completing unfinished routes if the game was saved after construction started but before the route was inaugorated (when road/railfinder is running).

I would say that this is quite common that AIs don't continue on half built routes if they are "stopped" by save/load at the middle of the process. CluelessPlus tries its best at detecting these half-built/broken connections and remove them while loading, which should get rid of stations/vehicles but not the road and probably not the depots.

I don't think you need to have a footnote about this in the wiki as I don't know any AI that can continue on building a new connection after loading a game.

[Kogut]

An interesting idea, I updated info about AIAI.