Botting the WNBA Season...

[MrBlueNose]

With not much to bet on over the summer(especially as I'm not a fan of baseball), I decided to start betting on the WNBA. One problem: I know absolutely nothing about the WNBA. So I decided to write a program that would make my predictions for me, hence the phrase "botting" in the subject. This is really just a hobby for over the Summer, and a fun way for me to keep my gambling fix.

I've fed the program a lot of stats, and from that it will predict the exact outcome of said game, based on various factors. The key to the program is that I don't tell it how to come up with the predicted results - for example I don't say "Hey, the home team wins 75% of the time so factor that into your results". I let the program try and figure out an exact percentage for itself, readjusting itself as more data comes in.

Do I expect this bot to be a winning bot? Most likely not. It's not very complex at the moment, and has a long ways to go. For example it doesn't factor in any players yet(although that's possible in the future). Mainly it's a way for me to bet all Summer on the WNBA, and for me is a lot more fun than just flipping a coin.

Anyhoo - the bot went "live" - as in I started "trusting" its picks on Friday. I've actually posted these on another forum(if someone wanted verification that I was doing it pre-game), and I'll start posting them on here. I should advise that I'd hope no-one trusts the bots picks religiously [img]/images/graemlins/wink.gif[/img] Anyway I'll get everyone up to date with the results so far...

June 15th:

New York +2: Prediction is New York to win by 1 point. They lose. Loss.
Connecticut +8: Prediction is Connecticut lose by 5. They lose by 3. Win.
Phoenix +10: Prediction is Phoenix win by 10. No bet.
Houston +11: Prediction is Seattle win by 12. Win by 13. Win.
LA +1.5: Prediction is LA win outright, by 4 points. They win by 9 points. Win.

Record:

3-1-0 (75%)

June 16th:

Indiana +3.5: Bot says Indiana lose by 3. They win by 10. Win.
New York +5: Bot says New York lose by 2. They lose by 8. Loss.
Minnesota +1.5: Bot says LA win by 6. They lose by like, a lot. Loss.
Houston +14: Bot says Houston lose by 11. They lose by 12. Win.

5-3-0 (62.5%)

And today, just the one game as Chicago took on Connecticut, Chicago at +8.5. Bot says Chicago would actually win by 2 points...and they win by 13 [img]/images/graemlins/smile.gif[/img] Win.

6-3-0 (66.7%)

And after writing all that...it predicts Detroit will beat LA by 2 points tomorrow, and the line is Detroit -2, so no bet tomorrow
[img]/images/graemlins/frown.gif[/img]

[Mr.JR]

Nice sample size.

Would love to see the results at the end of the season.

[fun160]

[ QUOTE ]
Nice sample size.

[/ QUOTE ]

Nice encouragement. The longest journey begins with the first step, no?

[MrBlueNose]

[ QUOTE ]
Nice sample size.

Would love to see the results at the end of the season.

[/ QUOTE ]

Haha. Yeah - I am completely breaking all "rules" when it comes to sample size [img]/images/graemlins/smile.gif[/img] Thats okay -as I mentioned it's really just a fun hobby, and something I always wanted to do. If I was really 100% serious about it at this stage I'd be plugging data from years ago, and running sims from past games and fine-tuning it based on results in the past. It's really just a work-in-progress, and hopefully something that might actually become a serious(and profitable) project in future years. I just figured I'd start posting results as I go, and update on programming changes and look at trends as the season goes on. Plus that way people can see it progress.

[TomG]

Sounds like a fun project. Posting your picks here is a good way to keep track of your W/L, but I know I personally would enjoy hearing about the various observations and tweaks you make to the model along the way. Please post as much or as little information as you would like to share. Good luck!

[rjp]

Hope this works for you, but consider predicting the odds. For example: even the worst of teams (in most situations) shouldn't be predicted to win by more than 10 points a majority of the time. I know WNBA teams will get trashed from time to time, but you're going to have a tough time predicting this result over the long haul.

[smbruin22]

i appreciate the picks.... and there's so little objective info (i.e. NO sagarin) out there about WNBA that i'd be inclined to give sample size some leeway.

[bkkdude]

pick the teams which are the most unattractive i bet they the good teams.....lol.

[MrBlueNose]

[ QUOTE ]
Sounds like a fun project. Posting your picks here is a good way to keep track of your W/L, but I know I personally would enjoy hearing about the various observations and tweaks you make to the model along the way. Please post as much or as little information as you would like to share. Good luck!

[/ QUOTE ]

I'll mostly be tracking the W/L via the program, because I'll write some stuff that'll provide a lot of detailed output other than just the W/L record, but I just like sharing this stuff on a public forum for a variety of reasons. It helps me track my progress I find writing it in public, rather than doing it all in private. And I'll definitely share any observations and tweaks as time goes on. Right now I'm probably going to just let it run for a month to see how things go and make sure there's no major hiccups, then analyze the data.

[ QUOTE ]
Hope this works for you, but consider predicting the odds. For example: even the worst of teams (in most situations) shouldn't be predicted to win by more than 10 points a majority of the time. I know WNBA teams will get trashed from time to time, but you're going to have a tough time predicting this result over the long haul.

[/ QUOTE ]

This is something that I'm looking into. I'd eventually like it to output a lot of different data, also in relation to the over/under, so I can determine where the best value is.

[ QUOTE ]
i appreciate the picks.... and there's so little objective info (i.e. NO sagarin) out there about WNBA that i'd be inclined to give sample size some leeway.

[/ QUOTE ]

Yeah I think there's enough people out there that like to bet no matter what, which was another reason I (a) wrote this, and (b) posted it [img]/images/graemlins/smile.gif[/img]

The line tonight is Sacramento Monarchs -4 vs Chicago Sky. Bot predicts Sacramento to win by 5 points, so Monarchs is the bet that the bot will be taking. Interestingly, if I tell it not to consider home/away records at all, it comes up with exactly -4 for Monarchs. I like when it predicts the spread almost exactly - feels like I am on the right track.

[rjp]

Let me also add that, from my perspective, the point spread is a secondary. I'm much more concerned with the team's probability of winning. Once you've got this the point spread can be approximated fairly easily.

The only problem is that when you're trying to gauge the successfulness of a probability you can't use simple W-L records, so you'll have to use some sort of loss function that can be difficult to interpret when the probabilities vary.