New AI competition

Discuss the new AI features ("NoAI") introduced into OpenTTD 0.7, allowing you to implement custom AIs, and the new Game Scripts available in OpenTTD 1.2 and higher.

Moderator: OpenTTD Developers

Post Reply
User avatar
DerekCZ
Engineer
Engineer
Posts: 105
Joined: 29 Jul 2014 18:08
Location: Czech Republic

New AI competition

Post by DerekCZ »

*TRIGGER WARNING: my grammar*
Also, I can not add more than 3 attachments, so look in the next post for remaining 2 savegames.

So let's start by making up some conditions that an AI has to fulfill to be able to compete in this semi-official competition;

• AI has to be able to handle at least 1 type of transport (so no Tree AIs)
• AI has to be able to handle it's type/s of transport effectively (so no eye candy AIs)
• AI has to be able to work without any additional NewGRFs (idk if there's an example, but whatever)

Ok, now let's pick our competitors (in alphabetical order):
[+] Spoiler
AdmiralAI
AIAI
AroAI
BorkAI
ChooChoo
Chopper
CluelessPlus
Convoy
CPU
FastPTPAI
NoCAB
OtviAI
PathZilla
Rythorn Airline AI
ShipAI
TransAI
That should be enough for this competition

Ok now that we have our competitors, let's set some rules; I want many very fast games, so
• map size will be "only" 512x512
• game duration will be "only" 5 years (3000 - 3005)
• vehicles never expire
• aircraft speed ratio set to 1/3
• PAX cargodist symetric
• mail cargodist symetric
• valuables cargodist asymetric
• other cargo cargodist manual
• temperate climate
• trains and ships can not go 90°
• airport noise stuff OFF (so 2 per town I guess)
• aircraft crashes Normal
• and some other settings, but keep in mind that good AI should adapt to all rules muehehe
• if an AI crashes or freezes, I am not going to restart it, it will have to hope that what it did before it crashed/freezed will be enough to compete
• all AIs will have default settings
• there will be no NewGRFs

And how are we going to play?
1) randomly select 4 AIs
2) let them compete for 5 years without any influence from me (The Player)
3) after 5 years, 1 AI will proceed to the next round; one with the best company value
4) and just for the lols, I will personally select 1 AI with the best aesthetics or something worth mentioning
5) after the round ends, I will export a save file and then attach it here so you can check it out by yourself if you want to

So, now that we are set, let's get ROOOOOOIGHT into the NEECOMPETITION!

ROUND 1!!!

Competitors: ChooChoo, AdmiralAI, FastPTPAI, OtviAI

BEGINNING: ChooChoo decided to start simply with forest to sawmill trainline, quickly followed by coal trainline. And just like ChooChoo, even Admiral started with coal trainline. PTPAI is thinking real fast, it already has two trainlines, and it also does not think cheap, considering that both lines are maglevs. And guess what, it picked the exactly same route as Admiral, thus transporting coal from mine to power plant. Another line is going from farm to factory. And the last but certainly not least is OtviAI, which also went for maglevs, but unlike others, it decided to transport passangers, or at least that's what I think, because it very quickly ran out of money, and judging from it's console log saying "Panic: Out of cash!", it quickly realized it's mistake. We will see...

MID-GAME: And how ChooChoo started it also continues; 7 out of it's 9 trains are freight, but now it started to work on a bit more complex, dual-track maglev network with some nice intersections, so the time will tell, it still has about 2.5 years left. Admiral dropped his idea of freight train lines and focused on growing only 2 towns with many busses. Not the most profitable way to play, but maybe it is planing something? FastPTPAI does nothing. That's it. It still has only 2 trains and 2 lines. According to it's console it is trying to find a new connection, but it always fails when trying to build a depot. I don't want to be pesimist, but it looks like PTPAI is out. Oh boy oh boy, Otvi is going for it! After it's small money shortage at the beginning, it quickly made itself a solid foundation for the rest of it's networks, RVs to be exact. And just for the sake of a nice gameplay it also made some boat lines, so extra points to you Otvi! (not really)

ENDGAME: This is it, the end of Round 1! So let's sumarize it a bit. ChooChoo REALY expanded it's double-maglev network, it now spans over many cities transporting passangers all over the place, seemingly taking advantage of cargodist. In the end, Admiral decided to try setting up an airline, but sadly the round 1 is over now, so it pretty much made all of it's money from road vehicles. FastPTPAI tried to recover by building 1 more trainline from iron mine to steel mill, but just like in Admiral's case, it is now too late. It should keep it's fast thinking, but what is done, is DONE. And OtviAI, unlike others, kept building and planning and looking at the world from the top of the income graph. I would also like to mention that it has very good pathfinding when it comes to trainlines, it can also update it's plans when some obstruction appears in the way of future train tracks.

SUMMARY:
Place 4: FastPTPAI with $160,532
Place 3: AdmiralAI with $1,616,418
Place 2: ChooChoo with $3,709,874
PLACE 1: OtviAI with $4,087,170

And the „Nicest looking empire“ award I goes to ChooChoo, mainly becase how it can handle large rail networks and it's ability to use junctions and stuff.

ROUND 2!!!

Competitors: AIAI, PathZilla, Chopper, Clueless Plus

BEGINNING: And here we go again, round 2 just went off! AIAI decided to just jump right in with 2 airports and the maximum distance possible between them. Pathzilla seems to be sleeping right now, or maybe it is waiting for other AIs to grow some cities? Chopper does exactly what it!s supposed to do; it builds heliports. And Clueless seems to be going for good old bus lines.

MID-GAME: AIAI has no problem slamming 8 airports next to each other and filling them with largest capacity planes. Thanks to this strategy, it's value skyrocketed, and... lol as I'm writing this one of it's planes crashed leaving jammed runway and 800 dead people. GG. Pathzilla did something, but it's still less than I would expect, considering that we are playing on relatively small map and there should not be that much pathfinding to do. Chopper has 7 heliports, but only 3 helicopters, and Clueless, like PathZilla, is extremely lazy with it's 12 busses. Well, at least it puts smiley faces in names of it's stations.

ENDGAME: In the end, AIAI completely OWNED the game. And I mean it. I have a pretty large monitor, and even when I expanded the income graph all over it, AIAI's line made all other companies look like a straight line on the „0“ level. After AIAI hit the 200 aircraft limit and after building it's 70th airport, it moved onto trains and busses, but slowed down a lot. I can't blame it however, because while it was busy dominating the skies, it presented pretty much no competition to other, ground-based AIs (except Chopper). I would like to summarize both PathZilla and Clueless now; they behave pretty similar, they both made couple road connections, and put a few vehicles to it. Not enough to compete with AIAI or Chopper, but yea, they tried. And Chopper just bough more helicopters and heliports, that's it.

SUMMARY:
Place 4: PathZilla with $217,428
Place 3: Clueless Plus with $563,552
Place 2: Chopper with $1,152,838


- empty space to signify the difference -
PLACE 1: AIAI with whopping $81,781,620

And now it's time for the „Nicest looking empire“ award II, which goes to Clueless Plus. I also considered PathZilla, but it was not able to show off it's infrastructure. And AIAI also made some nice rails, but I can not forgive it's airport spamming.

ROUND 3!!!

Competitors: AroAI, BorkAI, Convoy, CPU

BEGINNING: AroAI did not wait and right after the start it started building mid-range road connections between cities and filling them with busses. BorkAI, on the other hand, decided to wait for Aro to build it's network and then it used Aro's roads to transport passangers with it's own busses. Convoy went for the same strategy as Aro, it also chose the same cities. However, Aro was faster so instead of building roads by itself, it used Aro's just like Bork. Needless to say, all of those 3 Ais share the same network now. That does not go for CPU, which built simple intercity network with 1 vehicle in a seemingly random city as far from other companies as it could. Let the game begin!

MID-GAME: Aro is still expanding it's road connections, now spanning all over the map. But BorkAI scrapped the idea of road vehicles and focused on airlines instead. I was expecting the same behaviour as AIAI in round 2, but Bork looks like it knows it's bussiness. Instead of filling all of it's airports with maximum capacity planes, it also considers the size of the city, so in small cities it just builds a small airport with a few small planes. Definitely better looking than AIAI. Convoy is seriously competing with Aro, and according to it's company value, it is more than succesful. And CPU is still being asocial, building intercity bus lines only where there are no other companies.

ENDGAME: Round 3 is over so let's look at them results! They are very interesting indeed (or not, it depends how you look at it). Aro and Convoy did pretty much the same, they both chose the same tactic (bus connection), even sharing some roads. Bork, who also tried building roads, decided that 2 road-focused companies are enough and it started airport-building frenzy. After building 70 airports it realized that it has to buy some aircraft, so it ended up with 72 airports and only 35 aircrafts. CPU completely lost this round. It just made couple intercity lines, but that was not enough to compete with aircraft-buing Bork so it gave up midgame.

SUMMARY:
Place 4: CPU with ONLY FRICKIN $2
Place 3: AroAI with $6,076,882
Place 2: Convoy $10,097,700
PLACE 1: BorkAI with $13,262,154

And now it is time for our good know award, which I wanted to give to AroAI, but after completely jamming many cities with busses, I have decided to give it to Convoy. Good job there.

And with no more waiting let's start up
ROUND 4!!!

Competitors: Rythorn Airline AI, ShipAI, NoCAB, Trans

BEGINNING: Rythorn Airline started exactly how you would expect it to start; it built couple of airports and couple of planes, now it seems to be waiting for some income so it can build faster and faster. Honestly, I expected ShipAI to sink (get it? Cuz it's ship? Huehueh), mainly because all other AIs who said that they can handle ships usualy failed right at the start (personal experience). But hey, Ship managed to build some ports and buyos right from the beginning! NoCAB is really taking it's time, scanning all towns and possible connections, and Trans decided that the way to go is to compete with Rythorn Airline by building 2 hub airports. This is going to be a very interesting round, because we never had 2 aircraft focused companies before. Let's do it!

MID-GAME: Rythorn Airline did all it could, but unlike old AIAI it did not spam airports and airplanes all over the place, it looks like Rythorn is actualy trying to think about it's strategy. Ship is nicely connecting cities near water. There already are couple oilfields but Ship seems to prefer passanger transport. NoCAB is dead. Sorry m8, but it looks like it is frozen, it is not even posting stuff in it's console log. Bye. Trans developed a nice aggresive strategy. First it saw Rythorn building airports, so it also built airports to compete with Rythorn, but then it noticed that Ship's value is going up, so it also made couple of boat connections. Right now it looks like that we will not see trains or road vehicles in this round. Interesting.

ENDGAME: Well, this was not interesting at all :/ Rythorn Airlines AI continued building ariports, airplanes and company value. ShipAI did all it could, but because it is not able to build artifical water infrastructure, it quickly hit the top of what it can accomplish. Trans decided to adapt Rythron's strategy, and it adapted it very effectively. And guess what! NoCAB, after 4.8 years of carefuly watching and learning other's tactics decided to very originaly build airports. Sadly, it is now too late to start, so NoCAB failed...

SUMMARY:
Place 4: NoCAB with $268,336
Place 3: ShipAI with $940,322
Place 2: Trans with $9,025,860
PLACE 1: Rythorn Airline AI with $15,542,080

And as the nices looking company I would like to choose ShipAI, becase how it sends it's ships by the coastline of inland lakes.

Okay. 4 rounds are over now, we have 4 winners, so I think it is time for the final round of this competition. Ladies, gentlemen, attack helicopters and all 60 remaining genders, prepare for the ultimate battle that will propably not be that cool as you expect it to be.

FINAL ROUND

Competitors: OtviAI, AIAI, BorkAI, Rythorn Airline AI

BEGINNING: OtviAI started by building 2 passanger maglev stations, but it looks like it has a really hard time finding path for the conenction, so let's hope that Otvi will recover. AIAI built some airports and airplanes, there really is not much to describe. BorkAI sleeps right now, and Rythorn repeats what it did couple minutes ago.

MID-GAME: Otvi is certainly not tryharding. Instead, it uses all means of transport, ranging from ships to airplanes. AIAI tried some road connections, but quickly returned to airplanes and spamming airports. BrokAI woke up but it is not doing much, it only bought 1 road vehicle and 3 airplanes. And Rythorn Airline is doing Rythorn Airline Stuff, aka building airports, buing airplanes.

ENDGAME: Otvi realized that it can not compete with airline-focused Ais, so it looks like it decided for mostly aesthetic tactic, connecting cities with maglevs and roads. AIAI believed in itself when spamming airlines, but it did not realize that the largest, most expensive plane is not very effective with 9 people onboard, which resulted in a big drop in their company value. Rythorn played slowly but steadily, in the end it actualy tied with AIAI, but we will talk about that later. Bork decided to rather copy Otvi, but not very effectively. At least it transported some mail.

SUMMARY:
Place 4: BorkAI with $1,760,314
Place 3: OtviAI with $2,990,716
And we have 2 winners, I repeat, WE HAVE 2 WINNERS;
Place 1: AIAI with $23,291,456 and Rythorn Airline AI with $23,014,590

And why 2 winners? AIAI won just because it spammed airlines and it reached the 200 aircraft limit. But Rythorn was able to compete with ONLY 15 AIRPLANES, literaly. That's why I think that both AIs deserve it here.





FINAL CONCLUSION:

• What did we learn in this competition?
Nothing, selecting an AI to play with is your subjective choice.

• K m9 but what AI would YOU recommend?
If you just want to casualy play, then I would recommend OtviAI, but remember that we have still more AIs to cover in possible future competitions.

• Nice, but I like to tryhard as f*k and I want some serious competition.
AIAI

• And the nicest looking AI overall?
I think that would be either ChooChoo, OtviAI or ShipAI if you want some nice populated lakes

• And when will there be another competition?
idk

• Where?
in this topic, I don't want to spam forum with useless stuff like this

• Trump or Hillary?
Miloš Zeman

So, thank you for reading, and good night. And if it is not night, then just go to bed and think about your life and why are you concerned about AIs. Spolygon out.
Attachments
ROUND1.sav
(373.87 KiB) Downloaded 197 times
ROUND2.sav
(483.5 KiB) Downloaded 215 times
ROUND3.sav
(435.79 KiB) Downloaded 211 times
User avatar
DerekCZ
Engineer
Engineer
Posts: 105
Joined: 29 Jul 2014 18:08
Location: Czech Republic

Re: New AI competition

Post by DerekCZ »

Ther you have them 2 files. Also, if you have any questions, feel free to ask.
Attachments
ROUND4.sav
(384.33 KiB) Downloaded 215 times
ROUND5.sav
(463.66 KiB) Downloaded 206 times
Wormnest
Engineer
Engineer
Posts: 117
Joined: 14 Jul 2013 12:33
Location: Netherlands

Re: New AI competition

Post by Wormnest »

Thanks for testing. It's always interesting to see the different ways AI's can get used and you added nice interesting descriptions.

A few remarks:
It's a good thing to see a late start year for a change. However by starting that late in combination with no NewGRFs and setting aircraft speed to 1/3 instead of the default 1/4 you really are giving AI's that prefer airplanes a huge advantage.

I personally prefer somewhat longer games (but of course that's a personal preference). AI's that start slower can sometimes keep on going steadily up while some fast starters can reach their top soon and then stop improving.

By using the default settings you're giving an advantage to AI's that start very competitive by default. While AI's that by default try to be a little more easy going might be less profitable. I personally would try to set their options to comparable competitive settings as much as possible.
Especially with a small map size not being as competitive and not building as fast may cause an AI to have less profitable spots to choose from.

For the rest I enjoyed reading your test so keep going!
User avatar
Zuu
OpenTTD Developer
OpenTTD Developer
Posts: 4553
Joined: 09 Jun 2003 18:21
Location: /home/sweden

Re: New AI competition

Post by Zuu »

Thanks for the detailed post about your AI competition you carried out.

As for settings, a good start is to configure general AI difficulty in the settings window to hard before adding the AIs to your game. Then you will use the hard setting preset provided by the AI author.
My OpenTTD contributions (AIs, Game Scripts, patches, OpenTTD Auto Updater, and some sprites)
Junctioneer (a traffic intersection simulator)
alfredocuaresma
Engineer
Engineer
Posts: 2
Joined: 06 Jun 2021 10:06

Re: New AI competition

Post by alfredocuaresma »

Thanks a lot @DerekCZ , this review has been so helpful to improve my futures games :)
If there is any other updated tests of TTD's IAs, i would love to read them.
alfredocuaresma
Engineer
Engineer
Posts: 2
Joined: 06 Jun 2021 10:06

Re: New AI competition

Post by alfredocuaresma »

Just an update on this thread.
I've been testing several IAs, and based on my experience the best performance are achieved by these:

1. NoCAB (3.650.000.000€)
2. AIAI (375.000.000€)
3. OTVI (311.000.000€)
4. BorkAI (139.000.000€)


Image
Post Reply

Return to “OpenTTD AIs and Game Scripts”

Who is online

Users browsing this forum: No registered users and 10 guests