The solution i added to the above is two inputs, one's a net that scores value of combinations. Eg. Mohole+Mohole UNMI scores high as a combination. The other is raw strength. Mohole+mohole+UNMI is a known combination. It just sucks. So it picks the other known combination (CSM+all the cards that go along with that)
Early version of an AI you can practice TFM against
Forum rules
Please DO NOT POST BUGS on this forum. Please report (and vote) bugs on : https://boardgamearena.com/bugs
Please DO NOT POST BUGS on this forum. Please report (and vote) bugs on : https://boardgamearena.com/bugs
-
AReallyGoodName123
- Posts: 7
- Joined: 01 January 2025, 21:12
Re: Early version of an AI you can practice TFM against
https://boardgamearena.com/tableview?table=886658940
Bot fund miner for me in the last gen for 20mc. Final funding is the worst flaws right now.
Bot fund miner for me in the last gen for 20mc. Final funding is the worst flaws right now.
-
AReallyGoodName123
- Posts: 7
- Joined: 01 January 2025, 21:12
Re: Early version of an AI you can practice TFM against
Good call. Internal scoring for awards didn't consider the last production phase. The logs state it was ahead with no predicted way for you to catch up in the lookahead. Then the final end game round of production happened. Fixed.
-
FrankJones
- Posts: 2441
- Joined: 30 June 2024, 00:24
Re: Early version of an AI you can practice TFM against
tfmbot does not start the game when I invite.
What am I doing wrong?
Disregard, I figured it out.
What am I doing wrong?
Disregard, I figured it out.
-
AReallyGoodName123
- Posts: 7
- Joined: 01 January 2025, 21:12
Re: Early version of an AI you can practice TFM against
Don't invite. Join. See the instructions to find and join. Looks like you're in a game right now though? Just go to it 
FrankJones wrote: ↑22 July 2026, 01:22 tfmbot does not start the game when I invite.
What am I doing wrong?
Disregard, I figured it out.
-
FrankJones
- Posts: 2441
- Joined: 30 June 2024, 00:24
Re: Early version of an AI you can practice TFM against
Well, I am impressed.
The tfmbot plays ...competently. There were a few silly plays that reminded me of the pitiful other bot, but it used Point Luna well. Point Luna is annoying, even when used by TFMbot.
It placed Commercial district on that 2-plant spot near the center instead of blocking my city area. At one point it started spamming standard project asteroids for no apparent reason while holding 10+ cards.
It did a good job on the awards, funding one early that I couldn't possibly win, and avoiding the other 4 which I had easily. It claimed its one milestone before risking my swiping it and getting all 3.
The tfmbot plays ...competently. There were a few silly plays that reminded me of the pitiful other bot, but it used Point Luna well. Point Luna is annoying, even when used by TFMbot.
It placed Commercial district on that 2-plant spot near the center instead of blocking my city area. At one point it started spamming standard project asteroids for no apparent reason while holding 10+ cards.
It did a good job on the awards, funding one early that I couldn't possibly win, and avoiding the other 4 which I had easily. It claimed its one milestone before risking my swiping it and getting all 3.
-
FrankJones
- Posts: 2441
- Joined: 30 June 2024, 00:24
Re: Early version of an AI you can practice TFM against
Its starting picks were solid.
Mistakes or possible mistakes, and notable plays:
1) Playing Geothermal power for seemingly no reason, leaving some dangling energy susceptible to being tapped. Granted, I only had 6 cards, and no power tags, so consortium was not an immediate concern and tapping was unlikely.
2) gen 2 draft it took designed micro over fuel factory. Debatable. Maybe the experts can weigh in. It had energy to use for fuel factory, and I think designed micro is iffy there against an opponent with a pile of titanium. Not likely a significant decision either way. It did have Olympus conference though, so, there's that.
3) Gen 3 was a tough one. It drew Ganymede Colony when I already had 3 Jovian tags, and research when it had Olympus. It hate-drafted Ganymede. (I spent the rest of the game looking for that card and we eventually got to the bottom of the deck). I was holding mass converter and was happy to get Research, especially when I later fell into Mars U. I am not sure why it bought Ganymede colony though.
4) It placed natural preserve on a 2-steel spot when I think the card draw spot next to the other card draw spot would have been better.
5) Gen 4 it drafted Deep-well heating over Birds. Questionable. Especially since it didn't even want Deep Well; didn't buy it.
6) Gen 4 it played Vesta Shipyard at full cost, a bad play against Saturn.
7) Gen 5 it did various things then played business contacts. Probably should play that first, 5 draws to get optimal aero or media group.
8) Move 241 is the bad placement of commercial district. In my opinion, anyway. Playing the card at all was bad. It had just drafted Mohole area in a game in which temperature was lagging badly. That was the better play, I think. The game ended up going 12 gens, and it started lighting money on fire doing SP asteroids.
9) Gen 6 it played an ocean on 2 steel instead of card. Probably a good move since it had alloys. Would it have done that without alloys?
10) Gen 8 it took Lava tube over capital. Bad choice. The only remaining volcano was in a bad spot; there was a great spot for capital. In the same draft it also took lava flows over a better card. There was only one volcano spot left, drafting lava tube and lava flows made no sense. It promptly played lava tube in a bad spot. It could have played capital nest to its commercial district and two oceans and one spot away from its other city
11) NOTABLE: It had a 31-14 lead in money production and did not fund banker. Well done. I ended up funding banker myself and winning it by a lot. How far ahead would it needed to have been to fund banker?
12) NOTABLE: It played decomposers and ecological zone and racked up big points on both. Well played. Despite my stealing microbes with ants. I had not yet played Ants).
13) Gen 8 it funded landlord with a measly 6-5 lead. Questionable, but it ended up working; it won landlord by a lot.
14) Way earlier in this gen it should have played restricted area probably on a card spot, and then acrivated it. Maybe it gets anti-grav instead of me.
15) Gen 9 it took archaebacteria over mine (when it has alloys). Questionable.
16) Gen 9 it makes a horrendous play, Aquifer pumping when there are 6 oceans already played. Meanwhile it still has not played Mohole area. This would later be exacerbated when it only got two uses of Aquifer Pumping, failing to use in a 3rd time, allowing me to play large convoy and get the 9th ocean. In the gen it played aquifer pumping, it used all its steel on the card and couldn't use steel for the ocean. Also, it has lake Marineris in hand.
17) Gen 10 with temperature still nowhere, it took Lunar beam as first draft but did not buy it. As point Luna, with zero energy production.
18) Same gen, 10, it can play plantation when oxygen is at 7. It should do this immediately before I do (I didn't, but I could have). Instead (MAJOR ERROR), it did a SP Greenery. Major blunder. That makes no sense whatsoever.
19) Gen 10 it finally plays Mohole area. A bit late.
20) Gen 11 it takes peroxide power over technology demo. Bad play. AT this point, the score is 78-77, though I had a lot of big cards.
21) Still gen 11, it places a greenery not next to any of its cities. Andwith alloyed steel, it is not placing the 9th ocean with aquifer pumping. It ended up paying 34 for two oceans with aquifer pumping. While holding Lake Marineris.
22) Move 591 is when it start inexplicably spamming SP asteroids for no apparent reason. It did 4 in a row. Score is still 89-89 at this point. Maybe it was trying to rush end game, but temp was way too low to try this. All it accomplished was my getting fish out one gen earlier.
23) Gen 12 draft it takes SF memorial over Immigrant city and media group. Very bad play. Then it takes Nitrite reducing bacteria, Still ignoring Immigrant city. It does not buy bushes Despite having decomposers, adv eco, and 2 psychrophiles.
24) Same gen, it could land claim a spot next to my city and 3 oceans that I stupidly left empty instead of putting a greenery there. It did other unimportant things, but I still snoozed on my greenery and it eventually put land claim there.
25) It did eventually play Ganymede colony for 4 points. Not bad after all.
26) It eventually played Lake Marineris with no oceans left, making the Aquifer pumping play even worse.
27) For its final play it did an SP greenery for 2 points when it could have played 3 of its cards for 3 points.
Mistakes or possible mistakes, and notable plays:
1) Playing Geothermal power for seemingly no reason, leaving some dangling energy susceptible to being tapped. Granted, I only had 6 cards, and no power tags, so consortium was not an immediate concern and tapping was unlikely.
2) gen 2 draft it took designed micro over fuel factory. Debatable. Maybe the experts can weigh in. It had energy to use for fuel factory, and I think designed micro is iffy there against an opponent with a pile of titanium. Not likely a significant decision either way. It did have Olympus conference though, so, there's that.
3) Gen 3 was a tough one. It drew Ganymede Colony when I already had 3 Jovian tags, and research when it had Olympus. It hate-drafted Ganymede. (I spent the rest of the game looking for that card and we eventually got to the bottom of the deck). I was holding mass converter and was happy to get Research, especially when I later fell into Mars U. I am not sure why it bought Ganymede colony though.
4) It placed natural preserve on a 2-steel spot when I think the card draw spot next to the other card draw spot would have been better.
5) Gen 4 it drafted Deep-well heating over Birds. Questionable. Especially since it didn't even want Deep Well; didn't buy it.
6) Gen 4 it played Vesta Shipyard at full cost, a bad play against Saturn.
7) Gen 5 it did various things then played business contacts. Probably should play that first, 5 draws to get optimal aero or media group.
8) Move 241 is the bad placement of commercial district. In my opinion, anyway. Playing the card at all was bad. It had just drafted Mohole area in a game in which temperature was lagging badly. That was the better play, I think. The game ended up going 12 gens, and it started lighting money on fire doing SP asteroids.
9) Gen 6 it played an ocean on 2 steel instead of card. Probably a good move since it had alloys. Would it have done that without alloys?
10) Gen 8 it took Lava tube over capital. Bad choice. The only remaining volcano was in a bad spot; there was a great spot for capital. In the same draft it also took lava flows over a better card. There was only one volcano spot left, drafting lava tube and lava flows made no sense. It promptly played lava tube in a bad spot. It could have played capital nest to its commercial district and two oceans and one spot away from its other city
11) NOTABLE: It had a 31-14 lead in money production and did not fund banker. Well done. I ended up funding banker myself and winning it by a lot. How far ahead would it needed to have been to fund banker?
12) NOTABLE: It played decomposers and ecological zone and racked up big points on both. Well played. Despite my stealing microbes with ants. I had not yet played Ants).
13) Gen 8 it funded landlord with a measly 6-5 lead. Questionable, but it ended up working; it won landlord by a lot.
14) Way earlier in this gen it should have played restricted area probably on a card spot, and then acrivated it. Maybe it gets anti-grav instead of me.
15) Gen 9 it took archaebacteria over mine (when it has alloys). Questionable.
16) Gen 9 it makes a horrendous play, Aquifer pumping when there are 6 oceans already played. Meanwhile it still has not played Mohole area. This would later be exacerbated when it only got two uses of Aquifer Pumping, failing to use in a 3rd time, allowing me to play large convoy and get the 9th ocean. In the gen it played aquifer pumping, it used all its steel on the card and couldn't use steel for the ocean. Also, it has lake Marineris in hand.
17) Gen 10 with temperature still nowhere, it took Lunar beam as first draft but did not buy it. As point Luna, with zero energy production.
18) Same gen, 10, it can play plantation when oxygen is at 7. It should do this immediately before I do (I didn't, but I could have). Instead (MAJOR ERROR), it did a SP Greenery. Major blunder. That makes no sense whatsoever.
19) Gen 10 it finally plays Mohole area. A bit late.
20) Gen 11 it takes peroxide power over technology demo. Bad play. AT this point, the score is 78-77, though I had a lot of big cards.
21) Still gen 11, it places a greenery not next to any of its cities. Andwith alloyed steel, it is not placing the 9th ocean with aquifer pumping. It ended up paying 34 for two oceans with aquifer pumping. While holding Lake Marineris.
22) Move 591 is when it start inexplicably spamming SP asteroids for no apparent reason. It did 4 in a row. Score is still 89-89 at this point. Maybe it was trying to rush end game, but temp was way too low to try this. All it accomplished was my getting fish out one gen earlier.
23) Gen 12 draft it takes SF memorial over Immigrant city and media group. Very bad play. Then it takes Nitrite reducing bacteria, Still ignoring Immigrant city. It does not buy bushes Despite having decomposers, adv eco, and 2 psychrophiles.
24) Same gen, it could land claim a spot next to my city and 3 oceans that I stupidly left empty instead of putting a greenery there. It did other unimportant things, but I still snoozed on my greenery and it eventually put land claim there.
25) It did eventually play Ganymede colony for 4 points. Not bad after all.
26) It eventually played Lake Marineris with no oceans left, making the Aquifer pumping play even worse.
27) For its final play it did an SP greenery for 2 points when it could have played 3 of its cards for 3 points.
-
FrankJones
- Posts: 2441
- Joined: 30 June 2024, 00:24
Re: Early version of an AI you can practice TFM against
Now I forget what I did. Join how? Where?AReallyGoodName123 wrote: ↑22 July 2026, 01:30 Don't invite. Join. See the instructions to find and join. Looks like you're in a game right now though? Just go to it
FrankJones wrote: ↑22 July 2026, 01:22 tfmbot does not start the game when I invite.
What am I doing wrong?
Disregard, I figured it out.
never mind, I see now.
-
FrankJones
- Posts: 2441
- Joined: 30 June 2024, 00:24
Re: Early version of an AI you can practice TFM against
TFMbot keeps activating search for life even after it has found one.
-
FrankJones
- Posts: 2441
- Joined: 30 June 2024, 00:24
Re: Early version of an AI you can practice TFM against
Okay, game 2 versus Tfmbot:
I only won by 7. That's very impressive by TfmBot.
Granted, I walked right into getting maximum plants demolished by a gen 1 Deimos Down (which I saw tfmBot gain from its prelude); I suppose I just wanted to see if it would do it.
TfmBot committed the cardinal sin as Helion: Funding Thermalist, because, heat. I had screed up by investing in my own heat, which should have bene useless. Tfmbot build me out by spending 20 on Therapist, when My heat was going to otherwise sit there uselessly, and when Helion Tfmbot should be spending its excess heat on stuff.
Overall, a very solid game by Tfmbot (and a very poorly played game by me as Ecoline)
I only won by 7. That's very impressive by TfmBot.
Granted, I walked right into getting maximum plants demolished by a gen 1 Deimos Down (which I saw tfmBot gain from its prelude); I suppose I just wanted to see if it would do it.
TfmBot committed the cardinal sin as Helion: Funding Thermalist, because, heat. I had screed up by investing in my own heat, which should have bene useless. Tfmbot build me out by spending 20 on Therapist, when My heat was going to otherwise sit there uselessly, and when Helion Tfmbot should be spending its excess heat on stuff.
Overall, a very solid game by Tfmbot (and a very poorly played game by me as Ecoline)