AI Logic is quirky

Thanks again for the great game IronMike, I am seeing too many matches where a boxer gets knocked down and still loses the match. I think the odds need to shift dramatically in favor of a boxer who knocks down an opponent for the rest of the fight. Typically if a boxer gets knocked down first, he only has about a 10 to 20 percent chance of winning the match. Knockdowns take out so much from a boxer and typically you are doing to be likely to get knocked down again in the same round with a little bit less likelihood of getting knocked down again after each round past the Initial knock down.

Also, I am seeing a lot of computer opponents with horrible records in the top tier. In fsct, in general you seem to have very few fighters that have high winning percentages even with computer champions.

Just thinking a few tweaks on those two items might help with the realism some, with that said, this is by far the best boxing simulator I have ever played.

We’re going for old school records rather than the focus on the"-0" but we get that the balance sometimes looks out.

Agree, if that’s what you’re saying, that any knockdown (first or otherwise) should have greater effect than at present on outcome of the fight… Not sure of the stat that says first knocked down has 10-20% chance of winning. The question with that stat is what win percentage chance did the knocked down fighter have before the knockdown? I suspect it’s not actually that much different…

1 Like

I didn’t explain myself well. I was saying that the first boxer to get a knockdown in a match should have a much higher chance of winning the match. I have seen a few matches where the winner of the match on a decision had been knocked down once or twice. Typically in the real world, that would not be too common.

That we would all agree on :slight_smile:

Sometimes I intentionally am working a long term strategy though and I’m vulnerable early. Like I sometimes put it on the body, then pour it on when they gas. And it’s not uncommon for me to get caught once or twice before the body shots catch up to them.

For me, to be honest, I seem to do best just going with the natural ability. Maybe a good strategy is to go with natural ability early and then adjust as you get deeper in the fight. I believe there is only one true defensive strategy and it has not really done much for me. The WTBM strategies were fairly simple but they did tend to make a difference especially if your boxer is tired. I probably just need to play around more with the strategies but right now leaving on natural works fine for the most part.

It’s odd because the effects of the new instructions should have far more effect than those in WTBM. Interesting.

As a frame of reference I have played a total of 5000 matches of WTBM and 2000 of Boxing Manager to date and I am at roughly the same winning percentage of winning 80% of my matches. I used the wtbm strategies but for boxing manager pretty used no instruction. Part of it might be more that I typically have my stable of boxers look for opponents where they can have at least a 60% chance of winning. No instruction does not seem to negatively affect my gameplay but obviously others might have other opinions.

OK. One thing of interest might be the way we are starting (using stochastic methods) to implement fighter ai. so that the so Monte Carlo sims between rounds and iterate ai strategies like so :slight_smile:

That would really be cool. I am actually a Data Scientist and that is part of the reason I geek out over stats. I enjoy building out data models in Python and its different cool libraries. Definitely utilizing some advanced ways to improve the AI will make the game remain fresh.

1 Like

Agree about game freshness.

In this instance the Monte Carlo result size is dependant on the trainer ability… Thus “better” trainer bigger model, better recommendations of instructions…Your own experience also increases MC size to give the manager some impact/insight.