July 2025 - Losses, Environments & Timelines
Losses -- The only time shit gets sharper
Losses
July is the first real bad month since the re-tweaking of the goals and changing industries completely.
The first real tough losing period, both from a pure cash perspective and also a breaking of momentum side of things.
March, May and June had been awesome, even if you include the negative month of April which wasn’t horrific but still down, momentum was getting built, but this month… Big suck.
I don't mean the odd losing day or even losing week, I mean losing substantial money over the course of multiple weeks whilst still making the correct bets 98% of the time.
Previously November last year or January this year we had pretty substantial losing months, but was mainly down to there actually being clear bugs or logic issues in the model, essentially because I’m not a developer trying to do developer shit.
This, albeit frustrating, wasn’t that bad as it meant there wasn’t anything “wrong” with the model itself, it was just the sucky coding abilities, and obviously it sucks to lose a bunch of money but this falls into the “cost of doing business” type of bucket.
This time, the fucker that is July, (I think) that's not the case.
This just feels like one of those 1 in 100 type of month where every little thing went against you, every price or line, close game, injury, increased unit sizing edge, model variation, minor bug etc.
Culminating in some very rough days, some 1 in 50 type bad MF of a day, that realistically happens 2% of the time or 5-8 times a year if you are betting everyday (historically weirdly a Tuesday or a Sunday for me).
But this time, when you get 4 of these in a month, across all sports, you have a MF of a month.
Timelines & Zoom out
But zooming out.
The real issue is to understand "is this just expected bad variance, or is there a substantial real issue?"
Basically “is shit broke?”
This is the key in general with everything moving forward, it's also probably one of the less common things in business vs extremely common on the betting side.
Business is a little more cause and effect, more open loops. Especially if you’ve been in business for a few years, you know the cause and effect relationship well.
Betting though, is fully closed (which is good for me personally) but it also means the key question of "is shit broke?" Can be tough to answer.
It's also just a human nature thing, losing is worse than winning is good.
AKA Loss aversion.
Which I thought was funny that when I googled Loss aversion to grab an image, the first one is one that shows why this is so prevalent in betting.
Imagine this, every day, times by 1000x.
It’s tough, which is why zooming out helps.
I think this whole loss aversion thing is why gambling addiction exists at all but that's a rant for another day.
A bad day doesn’t really bother me nowadays, even a bad week not too much, the only one is a bad month as in-theory variance should even out enough that in that month you shouldn’t really lose if you have enough volume (number of bets per month) and edge (ROI per bet).
So if I make 100 bets per month at a 10% ROI/Edge, ($1 bets) that would be;
Expected profit: 100 × $0.10 = $10
Standard deviation: √(100 × 0.99) ≈ $9.95
Using the normal approximation (valid for 100 bets), the probability of losing money is the probability that your total profit is negative.
This equals: P(Z < -10/9.95) ≈ P(Z < -1.005) ≈ 15.7%
<took to Claude to create the above + these graphics>
Obviously nothing is truly 0.0%.
But for the purpose of this, if you’ve created a model that you think has a 20% edge and after 100 bets you are negative, the likelihood of that is only 1.5% (due to variance).
BUT, the likelihood that you’ve fucked up your model and overfit the shit out of it to trick yourself into thinking it’s good….. Is higher than 1.5%…. Speaking from experience.
And if you get to 200 bets at 20% edge and you are still down then that’s only a 1 in 1000 chance.
This is also why when I build models I aim for higher ROI and lower bet volume as mentally I think it’s easier to handle losing on 50 bets and thinking this might not be right than aiming for a 5% ROI and not being sure until you are in that 500 bet volume range.
But, this also does help once you get some volume of bets, as you can be more and more confident in your modelling. Meaning you can safely increase unit sizes if you match up to these performances.
Anyway.
When things do go bad, I to try counter these bad weeks or months, by zooming all the way out.
In my case I'd say it's been only about 7 months of "doing" sports betting this way.
There was years and years of previous specific knowledge of sports and betting principals and business knowledge that definitely helped, but the pure machine learning type creation of sports betting models only started in late October 2024.
And in reality the real progress only started to ramp up in Feb 2025.
So whichever timeline is used it's been 6-8 months of full real effort in this new method/business/project whatever you want to call it.
The speed of progress and profit in that time, including multiple weeks of very little, when travelling or moving around/house has still been impressive and promising.
By zooming out and asking would you have taken this profit/output during the first 6 months when you started? Even including any negative runs, bugs and all, the answer would obviously be yes.
Not from the money POV as I've had business that have been a lot more profitable than this, but those took multiple years to get to those points, instead the more promising point is how quick this has materialized.
By zooming out it helps to just put things into perspective.
At least for me personally.
It also helps to improve models in the future as well.
In the past the only times that these have been improved is when shit goes badly.
As when everything is going well, the saying of "if it ain't broke" comes to mind. But it's also pretty true.
For me, when stuff is going badly, massive improvements get made.
This month was no different.
That’s not to say that I’d rather have learned the lesson without the expense but everytime a drop happens the bounce back seems to be equal or better, and in theory you just run out of stuff to go wrong.
Losses are the only time when models get sharper, as by definition I wouldn’t be betting them if I didn’t think they were sharp….
Remember the goal.
As much as I’d like to have 20%+ ROI models for tons of different sports or 10%+ with crazy volume, in reality it’s unlikely, but I only need one or two.
You only need to be right once.
Apparently attributed to Mark Cuban relating to business, but it holds here.
If only 1 or 2 sports worked out after all this, those 2 would be so elite that it would pay for all of the fuck-ups of the rest….
Environment
Just a quick note on the environment impact of this.
Your physical environment is a massive predictor of success or failure.
Everyone knows this in the “the 5 people you spend the most time with sense”.
Or even how your postcode is more correlated to your wealth than your IQ.
But the environment in the purest sense of the word becomes valid here.
This bad run happened when I was back home (Cardiff), staying with the parents for 3 weeks waiting to move to a rental place in Bristol…
Correlated.
Obviously not causative.
It’s not to say that environments create results but there’s something to it here, I’m pretty sure I would have lost less this month had we still been travelling or back set-up in Bristol.
Obviously impossible to tell, but I have a hunch.
Travelling around Europe, working in Bristol, having great months, or going “backwards” and having the worse month of the year, even technically the 2nd worse month ever (in absolute terms, not relatively speaking though) when “back”….
Seems odd.
Anyway.
Future Stuff
I’m going to try write twice a month moving forward.
One will still be sports betting, niche related topics, with updates about the journey towards the goal and this switch of industries. Betting volume is still very low at the moment, but come late September and Q4 volume explodes so that’ll be a very interesting time.
The other post will most likely be about a concept or psychological theory that I’m trying to think about, might be interesting, might suck, either way, more to come.