On May 13 2020 06:44 starithm wrote: Awesome! I'm going to try out the Ranked Play tonight! :D
Please report back how you found it!
My record yesterday was 3-3. :O I did ZvZ well (except I got 4 or 5 pooled by a Zerg bot named Chris Coxe). I got owned by two protoss bots named BananaBrain and McRave. :O They were pretty good! I'm going to keep using SCHNAIL for now to get to my own ladder level 8~9 months ago when I was playing the game consistently.
Glad to hear that! One of the goals is to have it as a practice tool - if it fills that purpose, I'm very happy!
We did some showcase with CPL coach Neblime yesterday! We evaluated SCHNAIL, and the bots as a whole as a training tool - it was very entertaining, and you can check it out here: https://www.twitch.tv/videos/622027813
One thing is clear: People play MANY more practice games than ranked. I encourage you to play Ranked, the bots will be better evaluated by you doing so
On May 26 2020 04:50 Sonko wrote: I added a leaderboard to the website, you can check it out here: https://schnail.com/#/leaderboard
One thing is clear: People play MANY more practice games than ranked. I encourage you to play Ranked, the bots will be better evaluated by you doing so
Awesome. It's time for me to crush these bots in ranked matches. :D
On May 26 2020 04:50 Sonko wrote: I added a leaderboard to the website, you can check it out here: https://schnail.com/#/leaderboard
One thing is clear: People play MANY more practice games than ranked. I encourage you to play Ranked, the bots will be better evaluated by you doing so
Awesome. It's time for me to crush these bots in ranked matches. :D
If you're below 1900+ MMR, you might get embarrassed.
On May 26 2020 04:50 Sonko wrote: I added a leaderboard to the website, you can check it out here: https://schnail.com/#/leaderboard
One thing is clear: People play MANY more practice games than ranked. I encourage you to play Ranked, the bots will be better evaluated by you doing so
Awesome. It's time for me to crush these bots in ranked matches. :D
If you're below 1900+ MMR, you might get embarrassed.
On May 26 2020 04:50 Sonko wrote: I added a leaderboard to the website, you can check it out here: https://schnail.com/#/leaderboard
One thing is clear: People play MANY more practice games than ranked. I encourage you to play Ranked, the bots will be better evaluated by you doing so
Awesome. It's time for me to crush these bots in ranked matches. :D
If you're below 1900+ MMR, you might get embarrassed.
I'm embarrassed now.
I'm guessing you played some games? How did it go?
On May 26 2020 04:50 Sonko wrote: I added a leaderboard to the website, you can check it out here: https://schnail.com/#/leaderboard
One thing is clear: People play MANY more practice games than ranked. I encourage you to play Ranked, the bots will be better evaluated by you doing so
Awesome. It's time for me to crush these bots in ranked matches. :D
If you're below 1900+ MMR, you might get embarrassed.
I'm embarrassed now.
I'm guessing you played some games? How did it go?
I've been playing SCHNAIL bots since the beginning of May. But I keep losing to BananaBrain, Locutus, and Dragon bots.
On May 26 2020 04:50 Sonko wrote: I added a leaderboard to the website, you can check it out here: https://schnail.com/#/leaderboard
One thing is clear: People play MANY more practice games than ranked. I encourage you to play Ranked, the bots will be better evaluated by you doing so
Awesome. It's time for me to crush these bots in ranked matches. :D
If you're below 1900+ MMR, you might get embarrassed.
I'm embarrassed now.
I'm guessing you played some games? How did it go?
I've been playing SCHNAIL bots since the beginning of May. But I keep losing to BananaBrain, Locutus, and Dragon bots.
We added a replay browser to the website - you can search, browse, and watch games played on the SCHNAIL platform (Yes, you can watch a replay in your browser!)
On June 13 2020 10:36 xdq28635 wrote: Great work! I tried, it's fun. Thank you for the contribution. Question: how to fix the mouse acceleration issue in game ?
Thank you!
Mouse acceleration in w-mode is dependent on your host OS (Windows) mouse speed, so try to play around with those settings, and it'll probably be better.
On June 13 2020 10:36 xdq28635 wrote: Great work! I tried, it's fun. Thank you for the contribution. Question: how to fix the mouse acceleration issue in game ?
Thank you!
Mouse acceleration in w-mode is dependent on your host OS (Windows) mouse speed, so try to play around with those settings, and it'll probably be better.
The option is "enhanced pointer accuracy" or something like that in Windows, under the toggle for overall sensitivity, if memory serves correct. It needs to be off.
This is amazing! Are all the players on the leaderboard the bots only? Does playing ranked actually help the bots improve their skill? You're saying some of the AI are 1900+?
On June 14 2020 05:14 GGzerG wrote: This is amazing! Are all the players on the leaderboard the bots only? Does playing ranked actually help the bots improve their skill? You're saying some of the AI are 1900+?
Well thank you, I really hope you are enjoying playing SCHNAIL!
In this leaderboard ("AI leaderboard") they are: https://schnail.com/#/leaderboard - these are vs. human scores. The rating is standard ELO, so it's not directly comparable to human MMR, just evaluate it in proportion. The page linked has the info about how it is calculated and such.
And to answer your question: Playing ranked helps a LOT! - Indirectly just by inspiring the authors to be better, and providing replays to analyze their bots' performance. This kind of data is hard to come by. This is true for practice games as well - Directly since some bots have actual learning built it. This is currently only handled locally - as soon as we set up globalized handling of learning data, I expect some bots to have better overall ratings.
Comparatively few people actually play ranked games - I encourage everyone to try! Suprrisingly, the average human rating was lower last time I checked than the average bot rating!
@Jealous: Wow, I didn't know that, I have to test!
I’m having fun with it, but I think they’re way to easy to defeat with basic turtling that a near beginner could pull off. For example, with Zerg you can easily 1 base the hardest protoss by just spamming sunkens at your choke. They will not build air units to counter it.
Once they’re done sacking the entire map’s worth of ground units at your choke, you can win with 1 unit capable of knocking out the scattered photon cannons on the map.
On June 14 2020 23:39 Jae Zedong wrote: I’m having fun with it, but I think they’re way to easy to defeat with basic turtling that a near beginner could pull off. For example, with Zerg you can easily 1 base the hardest protoss by just spamming sunkens at your choke. They will not build air units to counter it.
Once they’re done sacking the entire map’s worth of ground units at your choke, you can win with 1 unit capable of knocking out the scattered photon cannons on the map.
Scores do not reflect your statement - average human ELO is lower then the average bot ELO. How many bots did you play, in how many matches? Once defeating one bot does not hold any statisticial significance.
"way to easy" is not really quanitifiable. I'm honestly curious about the level of play you were expecting, since the ELO scores are public.
Also, let me quote from the SCHNAIL FAQ:
Who writes the AI available on the platform? Talented hobbyists, students, and sometimes companies. A lot of bots were taken from the Student StarCraft AI Tournament database, but since that is an AI vs. AI tournament, and SCHNAIL is aimed at man vs. machine matches, we expect some divergence here. But anyone can write a bot, even with just basic programming skills - check out the tutorial on the SSCAIT page!
What's the MMR/ELO/rank of these bots? That is a surprisingly hard question to answer. So far, these bots have been competing against each other, and thus they are more tailored for a bot vs. bot meta, than a man vs. machine one. Their relative strengths can be gauged by looking at the SSCAIT ladder. Other than that, the general consensus by players who have played them that the best of them are around 1750 MMR / C rank.
(sorry for double post, pressed "quote" instead of "edit")
With all this said, don't take this the wrong way, I'm not flaming here - feedback is welcome, and I'm really glad you are enjoying the platform! I just feel that you were expecting something that were not promised at any point
EPT
