[Journey] Building an app that snipes football trends

A few months ago i posted here on the forum a journey of how i was growing my TG group about sports betting.
In one of the last updates i mentioned that to monetize the group my ideas were either to sell promotions to tipsters, betting related platforms or someone who could add value to the members, or if maybe would be better to find a way to monetize it myself, and i also mentioned that i was just playing around, programming a bot in python that "snipes" soccer opportunities based on statistics and many variables.

The thing is, between the curiosity i have to learn new things (coding, ML, scraping... ), my interest in sports betting and the ambition to keep growing, experimenting and trying to reach a successful goal, what started as a basic python script has become the beginning of a project i never thought i could build on my own....
I literally did it all by myself, reading a lot of course, and with help from AI, also through tests and errors and countless bugs; it was a real headache getting to this point...i swear there were days i had nightmares about this.... i could even say i became obsessed with the project because i saw myself taking one step forward and three steps back, and finally i achieved it...

Tbh at first i was scared because i've abandoned many projects i started in past due to poor results that caused me to lack motivation, but this time i feel it's different because there are INCREDIBLE results and for the first time i feel that extra push of motivation seeing that something i dedicated so many hours to is working even better than i expected...

To give more context, i'm currently running the project locally since it's still in "beta" and i have to adjust things and continue testing before launch it... The thing is, for about a month now, i've been sending only the high-probability trends that the app snipes to my TG group every day (some days there were no matches or simply no high-probability snipes).

The app currently scan 45 leagues, with around 950 teams and the markets it snipes are:

- Over 1.5 goals
- Over 2.5 goals
- Both teams to score
- Over 5.5 corners
- Winning streak
- Losing streak

While the plan is to expand to many more markets in the future, such as yellow and red cards, normal shots, shots on target, and other markets, for now it only snipes these markets to make it simple. The sniper script has almost 3K lines which take into account many variables to detect when a team meets a trend or when both teams meet that trend, and is based on a bunch of stats.
I'll leave a screenshot here to show an idea of what the snipe window would look like to a normal user, although im logged into my admin account and it will look a little different:



(I'm not going to blur the branding name since i'll probably end up changing it anyway, and the domain is already taken, or so i saw a few days ago lol)

As i mentioned, about a month ago i started sending the high-probability trends (the yellow ones) to my TG group so members could try the app, since they knew i was working on something like this on my free time.

And because of the good results (some of them have combined many bets and even made some very good profits, literally one guy won 160€ with a 5€ accumulator bet...something i NEVER recommend doing because it's very risky and unlikely to end well...but there are several cases like that in the group, with several large accumulators ending well), and many have messaged me privately asking if I sell the app or access to the trends, so there's real interest right now, and that's what motivates me most to continue.

Until a week ago, the results were just "empty / useless" because i wasn't keeping track of them. They were clearly quite good, but I wasn't registering them. And then while i was looking for ways to finance my project (i'll talk about this in more detail later), i discovered that i would need to compile the results in the future, both to improve the sniper and to get funding, investors, partners, etc because they will only have a real interest if they see that it is something that works and that people are interested in pay for it.

So a week ago the sniper hits 2/3 wins and the next day 5/5, and the guy who helps me moderate the TG group told me i should start tracking the data because even though it might just seem like a good streak, it was time to start tracking the results for the future updates and MVP validation. So i created an automated excel spreadsheet where i'm tracking a lot of valuable data. Besides the trend snipes results, i'm tracking the average odds, % win rate, and even win rate based on the odds range, in addition to other data..

I decided to start tracking the results from the day i created the excel tracker after the 5/5 wins day, as it seemed like "cheating" if i had included results from previous days. I know it's not much data, but i want to show yall the beginning...



Yes, i also noticed that most of the bets it detects are BTTS (both teams to score), but comparing it with several rival websites I realized that they have the same issue, so I suppose it will be the easiest market to predict, and as long as it's giving good results it's not something that worries me.

So, after 16 bets, with a simulated bankroll of $1K and each bet of $10, the sniper has an 81% success rate, $45 profit and with average odds of 1.65 (I will improve the odds thanks to the data im collecting with the excel and several other ideas i have in mind) and there are still 2 bets pending for today, which let's hope creating this thread doesn't jinx it and we don't lose the 2 bets...lmao

Now....¿What should be improved, added, or removed?

The purpose of this journey is to gather ideas, listen to opinions, and above all, motivate myself and track the process.
I'll be posting updates of the journey when there's news, but for now, im just looking at how to fund the development of the improvements, and searching info about all the legal aspects i should take care.
On the marketing side, i don't have a problem since, besides the 1.4K members group i have a +1K subs channel and also i have other ways to acquire telegram targeted traffic that will be 100% interested in the product, it's a proven method, so it's the least of my worries and i think it's my strongest point in the project.

Btw i'll be reading all your comments.
Good luck to everyone with your projects!

##### IMPORTANT #####

This thread is not for seeking funding or promoting anything.
Therefore, i will only accept ideas and opinions in the comments related about the topic.
Anything unrelated to the topic will be reported to a moderator, i don't want a warning/ban because of anyone playing around.

this is a great journey, i been wanting to create something like this, but i dont know how to go about it, even my telegram channel is low now but i have made $1k from my telegram channel with just 165 followers, i need to grow more

Congrats on the success, why do you only use 1% of the bankroll per bet? Given such a high success rate, wouldn't it be better to increase to 2% or even a more aggressive 5% if you really fancied?

I wondered as well, if you had 5 in a day, maybe you could do a smaller bet on 5 x 4 leg accumulators. This would hedge some of the risk but if your model is really at 80% accuracy, should pay off...

norsutom said:

Congrats on the success, why do you only use 1% of the bankroll per bet? Given such a high success rate, wouldn't it be better to increase to 2% or even a more aggressive 5% if you really fancied?

I wondered as well, if you had 5 in a day, maybe you could do a smaller bet on 5 x 4 leg accumulators. This would hedge some of the risk but if your model is really at 80% accuracy, should pay off...

Click to expand...

Thanks mate!!
Personally, im tracking individual picks to calculate overall win rate, win rate by odds range, by league, and other stats that i'll show in the next update since i upgraded the excel tracker.
So i don't recommend placing accumulator bets, its risky imo, but users will be free to do as they wish.

Regarding why i use 1% per bet, i believe it's the most conservative and "safe" approach in the long run.
Although the screenshot i shared showed an 80% win rate, in the next update i'll show how after almost 100 bets, the overall win rate and average odds dropped slightly after a 1-2 day losing streak.

UPDATE

After 1 month of tracking data every day (some days there were not many matches or there were simply no high-prob trends) these are the results.
Using a $1K simulated bankroll and a flat stake of $10 per bet, after 119 bets the win rate dropped from 81% to 69% and the average odds dropped slightly from 1.65 to 1.60, giving us a profit of $104 in a month.
I also added two new boxes to track win rate info by market and by league, and a chart to see the overall profit.
This will help me identify which markets and leagues are performing best and worst.



Honestly, i think the timing of starting to collect data was bad because many leagues have been in playoff or promotion/relegation phases this month, and although in many matches that was an extra motivation for both teams, in others it has penalized us a lot since one or both teams in some matches no longer had anything at stake, so although there was a trend, the context was somewhat more relaxed than in regular league matches or where there is something at risk.
Anyway, i think these are very good results for a first month of basic tracking, so to speak, being all in "raw" format without having made any adjustments yet and taking into account ALL sniped picks regardless of whether they are low/high odds or the context of the match.

If we want to go a step further and see what the results would have been by adjusting the stake based on the "direct risk" using the odds as a reference, and using stakes from 1 to 3, it would look like this:

≤ 1.59 → Stake 3 ($30)
1.60 – 1.79 → Stake 2 ($20)
≥ 1.80 → Stake 1 ($10)

Final profit: $247
Or in other words, $143 more than using a flat stake.

For now, i won't adjust anything and will continue tracking data, although with the World Cup starting in a month, i don't think there will be many matches that aren't from lower leagues.
I'll also take this break to edit the frontend and add some features i have in mind.

See yall in the next update!!!

I saw your 19M deck post. I need RT amplification for tech/AI product launches (not crypto, not NSFW) Can we work something out directly?

Lovely update, to clarify, are the odds listed what you get from the bookies? Does your model just give a binary yes or no on the bet?

If so, assuming it gives a degree of confidence, i'd be interested in a formula where you bet more on the longer odds. E.g. If your model predicts that X is likely to happen, longer odds would be more profitable long term. However, if it simply reports that a bet is *more likely* to happen vs the bookies odds, then your current formula is probably correct