Arguing With Robots and Trying to Find a Bet

Moving from construction into what is basically data analysis for sports betting, I had little to no knowledge of computer programming. I knew how to use the things in front of me, but very little about what made them work in the background.

The way the companies I work for are structured, I didn’t really need to know.

Nevertheless, with the development of AI, I couldn’t help getting involved. You want to learn how to do something different.

And by “learn”, I mean making these fucking robots do things you have no idea how to do yourself.

You tell them what you want. They don’t do it. You argue. They sort of do it. You argue some more. Back and forth, here and there, until eventually you get something you quite like the look of.

The first thing I managed was a small website for a friend. I could probably have done that without AI, having made something similar before, but the experience was quite fun. It felt like I’d learned something.

Maybe not programming exactly, but how you need to organise and present things so they work digitally. It sort of rewires your brain for the next thing you try.

So, naturally, the next thing was betting.

I wanted to build something along the lines of the tools you see touted around the internet: something that finds value bets, or at least steers you towards somewhere worth looking.

At work, we use large Excel documents and models, checking things against each other. A dashboard like the one I’ve created isn’t really how we do it.

But I wanted something I could use myself. Somewhere to put all the bizarre things in my brain that I look for when trying to find bets still available on the table.

That’s where the Daft-a-Peth LOTSO dashboard comes in.

It isn’t the system used by the large betting syndicates we work with. They have their own quants and modelling, and that stuff goes way beyond what I’ve put together here.

This is more the everyday bloke’s version. Have a hunt around a normal bookmaker, get your fiver down, have a bit of a laugh and try to get a little profit back.

The syndicate work involves a lot of quantitative modelling, xG and then mixing that with what us humans actually see with our eyes.

All the things people argue about whether they make any difference.

Momentum. Bad luck. Playing well and not scoring. A goalkeeper playing fucking shit.

Things that might help explain why a team will or won’t win, and whether the price properly reflects it.

Because finding a bet isn’t just deciding who you think will win. It’s deciding whether the price is worth taking.

If I think something has a 50% chance, evens is the fair price. To have an edge, I want better than evens—or, if I’m taking evens, I need to think its chance is better than 50%.

Taking 4/5 about that same 50% chance gives the bookmaker the advantage. You might pick plenty of winners and still struggle to make money long term.

That’s really what I’ve tried to build: something to help me look for those differences.

Not every interesting number is a bet. It gives me somewhere to start looking, compare the prices and decide whether I actually fancy it.

So if I’m walking through town and passing Coral, I want to be able to have a look, find something worth investigating and slide in to stick a tenner on.

If I’m backing it at evens, I want a reason to think it has a better chance than the price suggests. Whether I’m right is what the long-term record will tell us.

I’ll be using the dashboard myself and putting up write-ups explaining what it’s found, what I’ve backed and why.

Then we can see how it gets on.

So, like many things on this site, this has had a bit of help from AI. I’ve got dyslexia and whatever else makes my brain refuse to concentrate properly on spelling and grammar.

These write-ups will be dictated into AI and rewritten, so save yourself the bother of putting them through an AI detector. Yes, I’ve used AI. It helps me sound like someone with a brain.

Soz. Not soz.