Jul 18, 2026 · 22 min · 6 segments
The post AI in Sports Betting: What Works and What Doesn’t appeared first on FTS Income.
The challenge with talking about AI and betting is that that conversation can exist at several different levels at once and then they tend to get mixed together in ways that aren't helpful.
Tasks that took hours can now take minutes.
Pattern identification required, either specialist software or substantial manual work can now be done with a prompt and a spreadsheet.
At a different level, the bedding content space like all these things can absorb AI as a marketing category, which means there's now a massive amount, huge substantial amount of content, services, products that will use the lines and the words AI to describe things that have always existed under a different name or do not work in what they claim or actively do harm to the people who rely on them.
And then at the third level, the one I want to focus a bit more on, there's specific ways that AI tools used without the right foundation can lead you in completely the wrong direction, not because the tools are broken, but because the people using them are perhaps using them at the wrong stage of the process for the wrong purpose.
So I think we need to look at that, as I say, in those three sections, what AI can do well for you, just looking for certain impacts of overselling of it and what harm so where does AI genuinely help the most honest way to describe what large language models are good at is processing and organising large amount of text and data quickly identifying patterns producing structured output they are not small capabilities they are specific ones that can be speeded up to seconds And so for systematic bettors with data, the genuine use cases cluster around three sort of areas.
The first being data interrogation.
If you have a record of several hundred bets, odds, stakes, markets, results, an AI tool can identify patterns in that data significantly faster and without any subject opinion compared to manual analysis, which odds ranges have performed best and worst, whether performance varies by market or... day of the week where your results diverge from what a level stakes approach would have produced compared to different staking patterns and that is work that used to require hours and hours with a spreadsheet or you'd have to employ someone with specialist software and now it can be done with a prompt and a few minutes work.
The second area is review framework construction.
So one really useful application is asking an AI tool to help build a structured review process for a system, what to measure, how often, build a baseline, measure against that baseline, what threshold should trigger a formal assessment.
The tool does not have your data and it cannot tell you whether your system has an edge, but it can help you think through the structure review of a process in a way that is more rigorous than starting from scratch.
And the third is sanity checking reasoning.
So before placing a bet or making a decision about a system, walking through that logic with an AI tool, stating your reasons, asking it to identify gaps or inconsistencies.
That can surface things that are easy to miss when you are close to the decision.
It's not asking AI for a betting opinion.
It's using it as a structured thinking tool.
So I did an AI course.
The first thing that I did with the guys was give it data to interrogate, let's find a system.
But then we took that and went, rather than going straight to betting and staking, we then asked AI, to basically interrogate and test that system, point out its flaws, explain why it might not work forwards, which is basically all we want, all we're interested in.
We can find anything historically, but will it work forwards? Why would it have a good case of working forwards? Why wouldn't it? So we built a structure around taking that and forward testing and basically saying, yeah, we've got this.
Does it actually work? What are the chances going forward of it working? So that sort of data interrogation and then the ability to let's put some reviews in place.
If the edge is X when we start and it falls below Y, we want to pause, we want to look, we want to try and identify Y.
And AI can help you do all that so much quicker than anything manually.
And as I say, a lot of it without, as long as you don't lead it without any subjectiveness, one of the other big dangers is leading AI.
The challenge with talking about AI and betting is that that conversation can exist at several different levels at once and then they tend to get mixed together in ways that aren't helpful.
Tasks that took hours can now take minutes.
Pattern identification required, either specialist software or substantial manual work can now be done with a prompt and a spreadsheet.
At a different level, the bedding content space like all these things can absorb AI as a marketing category, which means there's now a massive amount, huge substantial amount of content, services, products that will use the lines and the words AI to describe things that have always existed under a different name or do not work in what they claim or actively do harm to the people who rely on them.
And then at the third level, the one I want to focus a bit more on, there's specific ways that AI tools used without the right foundation can lead you in completely the wrong direction, not because the tools are broken, but because the people using them are perhaps using them at the wrong stage of the process for the wrong purpose.
So I think we need to look at that, as I say, in those three sections, what AI can do well for you, just looking for certain impacts of overselling of it and what harm so where does AI genuinely help the most honest way to describe what large language models are good at is processing and organising large amount of text and data quickly identifying patterns producing structured output they are not small capabilities they are specific ones that can be speeded up to seconds And so for systematic bettors with data, the genuine use cases cluster around three sort of areas.
The first being data interrogation.
If you have a record of several hundred bets, odds, stakes, markets, results, an AI tool can identify patterns in that data significantly faster and without any subject opinion compared to manual analysis, which odds ranges have performed best and worst, whether performance varies by market or... day of the week where your results diverge from what a level stakes approach would have produced compared to different staking patterns and that is work that used to require hours and hours with a spreadsheet or you'd have to employ someone with specialist software and now it can be done with a prompt and a few minutes work.
The second area is review framework construction.
So one really useful application is asking an AI tool to help build a structured review process for a system, what to measure, how often, build a baseline, measure against that baseline, what threshold should trigger a formal assessment.
The tool does not have your data and it cannot tell you whether your system has an edge, but it can help you think through the structure review of a process in a way that is more rigorous than starting from scratch.
And the third is sanity checking reasoning.
So before placing a bet or making a decision about a system, walking through that logic with an AI tool, stating your reasons, asking it to identify gaps or inconsistencies.
That can surface things that are easy to miss when you are close to the decision.
It's not asking AI for a betting opinion.
It's using it as a structured thinking tool.
So I did an AI course.
The first thing that I did with the guys was give it data to interrogate, let's find a system.
But then we took that and went, rather than going straight to betting and staking, we then asked AI, to basically interrogate and test that system, point out its flaws, explain why it might not work forwards, which is basically all we want, all we're interested in.
We can find anything historically, but will it work forwards? Why would it have a good case of working forwards? Why wouldn't it? So we built a structure around taking that and forward testing and basically saying, yeah, we've got this.
Does it actually work? What are the chances going forward of it working? So that sort of data interrogation and then the ability to let's put some reviews in place.
If the edge is X when we start and it falls below Y, we want to pause, we want to look, we want to try and identify Y.
And AI can help you do all that so much quicker than anything manually.
And as I say, a lot of it without, as long as you don't lead it without any subjectiveness, one of the other big dangers is leading AI.
The rest of this transcript — segmented and speaker-labeled, so you land on the exact moment something was said
Search every transcript — by keyword, by phrase, or by meaning, across every show Radar indexes
Trends — what is surging across podcasts, measured against its own baseline
Alerts — when a name you follow appears in a newly indexed episode
No account is needed to search Radar.
