- My Forums
- Tiger Rant
- LSU Recruiting
- SEC Rant
- Saints Talk
- Pelicans Talk
- More Sports Board
- Coaching Changes
- Fantasy Sports
- Golf Board
- Soccer Board
- O-T Lounge
- Tech Board
- Home/Garden Board
- Outdoor Board
- Health/Fitness Board
- Movie/TV Board
- Book Board
- Music Board
- Political Talk
- Money Talk
- Fark Board
- Gaming Board
- Travel Board
- Food/Drink Board
- Ticket Exchange
- TD Help Board
Customize My Forums- View All Forums
- Show Left Links
- Topic Sort Options
- Trending Topics
- Recent Topics
- Active Topics
Started By
Message
re: Sports Betting AI Algorithm Prediction
Posted on 11/12/25 at 9:36 pm to LSUTIGERS74
Posted on 11/12/25 at 9:36 pm to LSUTIGERS74
How are the picks graded as in what makes it A vs elite?
Prob and edge seem inconsistent across all grades.
This post was edited on 11/12/25 at 9:37 pm
1
Posted on 11/12/25 at 9:39 pm to Gings5
Page 2 second post should help understand the tier grading thresholds
Posted on 11/12/25 at 11:04 pm to LSUTIGERS74
Site don’t work
Posted on 11/13/25 at 1:59 am to LSUTIGERS74
The site is down.
Posted on 11/13/25 at 6:21 am to uptowntiger84
Site is back up, thats my fault I crashed it late last night trying to get picks to auto update with spread movements.
Posted on 11/13/25 at 6:23 am to LSUTIGERS74
quote:
Either way, it's a strong pick. The -5.5 line is objectively better for betting, even if it drops below the Tier A threshold.
You're going to have to explain this one again because it makes no sense at face value. The line gets better, but the pick gets worse (lower tier)?
Only way this makes sense is if your algorithm is only looking at lines and nothing else i.e a home fav un this situation is more like to cover when THE LINE is -6.5 than when it is 5.5.
Which, you have to admit is not logical.
Probably an area of improvement
Posted on 11/13/25 at 6:26 am to LSUTIGERS74
Nice. BTW I wasn't trying to come off like my free chatgpt prompts were better.
Ill post my slips in a bit from your site!
Ill post my slips in a bit from your site!
Posted on 11/13/25 at 8:00 am to castorinho
quote:
Only way this makes sense is if your algorithm is only looking at lines and nothing else i.e a home fav un this situation is more like to cover when THE LINE is -6.5 than when it is 5.5.
I completely agree with what your saying. I"m not sure what he's doing in detail with how tier A and B are selected, but I can see where -6.5 at -110 is the preferred bet over -5.5 at -150. The value of the bet is likely the reason why in this scenario the difference in Tiers, at least in my opinion to your question. No one wants to make single bets on -140 and -150 too often. You're looking for -120 and better.
Posted on 11/13/25 at 8:08 am to TeddyPadillac
Yeah my comment is based on the juice being the same, only the line having moved.
So - 6. 5,-110 vs - 5.5 - 110
So - 6. 5,-110 vs - 5.5 - 110
Posted on 11/13/25 at 8:49 am to LSUTIGERS74
quote:
Yes with the correct wording in a prompt
Grok is better than Chatgpt. Grok started kicking out crazy coding but i have it act like an MIT mathematician. But yeah they sometimes post games thatvarent playing
But so far its about 12-6 on the games i have asked about
Posted on 11/13/25 at 8:57 am to LSUTIGERS74
I was expecting some AI tool where you can search any matchup & it would give you a pick/rating.
Looks like it’s only about 20 picks
Looks like it’s only about 20 picks
Posted on 11/13/25 at 9:46 am to castorinho
quote:
castorinho
Yes it is completely counter-intuitive and I think I just figured out why. The algo takes the opening line value and creates a house probability based on about ~80 data points then compares that to market probability at that moment. So when the line shifts it keeps the house probability the same and recalculates against the market prob which just shifted up when the line went from -6.5 to -5.5. This is an issue I am working to fix now, I am very grateful you found this as it helps explain a few things as it shifted this pick down to tier B because our confidence_score (i.e. edge) is calculated from that diff. in house vs. market. The good news is you found an exploitable bug and changing my thresholds to capture this improved the win rate 5 points because it shifted 8 games from 2024 & 2025 seasons into tier A that went 8-0. Send me your email address and I will set up a free account for you as a thank you.
Also taking a look at another option to shift house prob with lines but still running backtest on that before I can deploy.
Posted on 11/13/25 at 9:49 am to JimTiger72
quote:
JimTiger72
This AI tooling is in the works but haven't gotten that far down the development path on add-ons. For now the algo takes all the games for the week in an initial run and creates a set of data for the full set, then my weekly pick generator tosses out games that have poor probability and edge (weeding out), then it takes the remainder of games that have a decent set up and grades them. That is why there is only 21 games even considered for a pick out of the ~52 games this week.
Posted on 11/13/25 at 9:59 am to LSUTIGERS74
Can you explain "edge" to me?
Posted on 11/13/25 at 2:12 pm to WITNESS23
quote:
Can you explain "edge" to me?
Our edge is not the typical edge model Vegas would apply and calling it edge is probably not the best term, we actually define it as confidence score in how it works section of website, but basically it is house probability - market probability. This is a bit of a paradox but it works really well in our backtesting win rate and tier selection because as the spread gets worse (bad for bettors) the gap between our house probability and market prob gets wider (market prob goes down with spread going up, i.e. less chance to cover). However if the house probability was already >60% and the edge jumps over our +25pp with the widening gap means the pick gets pushed into our tier A picks. A bit confusing but the data on those picks versus using an adjusted probability with line movement sent the backtesting wins way down. The best thing to do is take the picks as close to kickoff with Tier A (as long as the spread doesn't move it out of Tier A) as that uses the latest lines in the calc.
Posted on 11/14/25 at 8:14 am to castorinho
quote:
castorinho
Thank you for prompting me to look into the line issue, I was able to refine the algorithm by changing the threshold down to 20 / 10 (edge) which pushed borderline picks (just below Tier A threshold) into Tier A and improved the win rate across the full 2021 - 2025 backtest.
Loading Twitter/X Embed...
If tweet fails to load, click here. Tier A:
Record: 392-182 (575 picks)
Win Rate: 68.26%
Tier Elite:
Record: 153-51 (204 picks)
Win Rate: 75.00%
Combined (A + Elite):
Record: 545-233 (779 picks)
Win Rate: 70.03%
The other thing of note from the deep dive, was trailing the probability with line movement was not great for win rate as the trailing factor is not linear for points move greater than +/- 1 so that got thrown out the window, but the outcome of this was the backtesting shown above uses the closing line before kickoff as that is the last data point to pull in the historical data set. I ran opening line backtesting and the WR dropped to roughly breakeven versus what I show above.
This post was edited on 11/14/25 at 8:25 am
Posted on 11/14/25 at 8:09 pm to LSUTIGERS74
I know see I got anchored.
Posted on 11/15/25 at 7:28 am to LSUTIGERS74
Used 2 elites and 1 A. Let's get it!
Posted on 11/15/25 at 8:47 am to LSURoss
Hopefully this is legit.
I made a round robin with some of the “C” list which is supposed to be fades and Troy got destroyed 33-0. Didn’t start out too good.
I made a round robin with some of the “C” list which is supposed to be fades and Troy got destroyed 33-0. Didn’t start out too good.
Posted on 11/15/25 at 9:51 am to LSURoss
I don't do parlays on this system but would have done well last week starting 3-0 on Tier A. It isn't designed to take filter parlays but if you do, use home favorites only and Tier A games.
Page 3 of 4
Popular
Back to top
Follow TigerDroppings for LSU Football News