Statistics Tool [Tutorial 4] How to work on the Backtesting feature?

Statistics Tool [Tutorial 4] How to work on the Backtesting feature?

There's a lot of buzz around our spectacular statistics tool with the Backtesting and Spreadsheet.

Rightly so!

We are offering something never has been offered before—an ultimate tool for the traders, punters, tipsters and also for betting models. The notification system itself was already something new and fresh, something unique and the first one on the market! The notification system needed a long testing period to get satisfying results, sometimes months of testing.

At that time, we had already Backtesting system written on the dashboard, and recently we launched a Beta version.


With Backtesting statistics tool, the users can test their criteria based on the previous data, on the already finished games.


The notification system itself was already something new and fresh, something unique and the first one on the market!
The notification system needed a long testing period to get satisfying results, sometimes months of testing. At that time, we had already Backtesting system written on the dashboard, and recently we launched a Beta version.


With Backtesting statistics tool, the users can test their criteria based on the previous data, on the already finished games.
This allows cutting the testing period from months to a few hours only! We are working on getting the Backtest time period for even shorter and at the same time, to check a larger amount of games and ability to target.

 

Backtesting statistics tool

The idea of the Backtesting tool is simple actually. You create a model/notification and let it run through the database of finished games.
 

You will get the results into the Backtest Spreadsheet and can see how the particular model would have performed. Yes, this doesn't guarantee 100% performance on future games, but this gives a good indication for different methods.


To find the holy grail, the method which suits, you can test and try different paths. I prefer to get as big as possible database from the Backtesting system to start.


The whole conception is not to get good enough results into Backtest Spreadsheet with the first shot. Take the Backtest results as the canvas for your masterpiece! It's a start, not the finish.


I have managed to get good enough results on corners, first half under and overs, full time under and overs. Betting on the outsider, but not yet going on the favourite.


This shows that there's no wrong or correct starting point; the sky is the limit.

Asian V bookie bets

Why is this important and why I'm trying to make also notifications/models on big leagues only?

I made my first bets around 12 years ago.

Of course, I didn't get a flying start, and I didn't have any issues with accounts. At some point around 6-7 years ago, I stumbled on a big block. My accounts were limited to a few euros maximum stakes in all of the major bookies we all know. I can take a screenshot with 5 euros maximum stake on Premier League games.

That's where the fun stops!


Some new bookies are arising now and then, but most of them don't offer the games our favourite bookies are offering. So as it stands, I'm left with two options. Either build some excellent models where I'm not able to stake more than a few euros or create betting models on Asia bookies where I can bet, but only on selected leagues. 

With our fabulous tool, so far, I have managed to get around 5%-10% yield on Asian only leagues. Around 5%-7% Yield on major leagues only, and believe me, this is a very good result.


Why so? 

The main two differences are the limits and being limited and the offer of different leagues and competitions.
The most convenient way of betting in Asia is to use some agent services, like Sportmarket. There are some more similar sites, of course, but about the pros and cons of different sites, I will write in the next blogs.


I'll make some easy to follow comparisons:

Sportmarket: Much bigger limits - you will not get limited - much less leagues covered.
Bookies: Much smaller limits - you can get limited very fast - much more leagues covered.

This is an easy and short explanation of main differences. Of course, it's much easier to get a good Yield and profits on bookies on lower random leagues. But if you become too good for them, you need some alternatives and change the way you are betting.

 

Model creation using our statistics tool

As you can see, I started again from the spot where I have plenty of data to play with. 6474 alerts, profit is - 31000 and Yield is -5%. A big red minus in profits? Yes, fine by me, as long as I have a sample base and I will show you why!


Imagine this; you have a sample base of 400 games and - 4000 in profits. Who is in a better position to start finding something profitable? I am! This is also a perfect sample base to show how to use, and why to try out all the different filters.


In general, I suggest starting playing with ODDS and TIME filters.

statistics tool

We have odds range from 1.55 to 5.25. Happens and thats why I suggest to start from here. There can be some small mistakes in the odds, traders flipped the odds for few seconds for example. I will change the range from 2.5 to 5.25.


A loss of - 2577 and 209 games. I checked some of them manually, I mean the odds, and there some probable trader errors and in some cases, the goal were scored at the same time etc. Let's clear the filter and try again with 1.55 to 2.5.


Still 6293 alerts and -28k. Now, let's play with these odds ranges further more. 1.55 to 1.6 to 1.7.


Im looking to create a notification with good odds, so my hit rate don't need to be like over 70%, but I can get good odds always and this makes also higher chance to find the markets from different bookies/Asia. 1.9 to 2.1 and we are on a 3162 alerts and -14k in profits.

We cut half of the alerts, but we also cut half of the - and still have plenty fo alerts to play with!

statistics tool

Time filter

The current range is from 61 to 79. A small gap, because this is how I built this notification, used a stricter time frame. 

This filter here is very underrated.

I'll prove it.

A lot of people doesn't think that with this filter, we also can effectively change the submarkets, for example. Let's change the range minimum from 61 to 65, for example. We changed the starting time from 61 minutes to 65 minutes, AND from 3162 alerts and -14k, we moved to 1660 alerts and -4k. Statistics tool used in full force!

statistics tool

We lost a lot of alerts, but we also removed a lot of negative balance and still have plenty of alerts to adjust and filter. I remind you that this is not the only way to do things here, but I'm giving you examples and ideas on how to use the tool. 

Country and League

The next step is our usual last steps, for some users, the first steps. We are checking the country and leagues lists.

Working on them is purely personal, in my opinion.

I'm removing the leagues and countries which are not suitable for my timezone—also removing some smaller leagues where I wouldn't be able to stake on anyway.

What's the point to have a model with 30% Yield on paper when I cant get like 50% of the alerts? I'm not just clicking buttons on all the red leagues and countries also. It would be best if you still had sufficient samples in the leagues and countries BEFORE eliminating them from future notifications.
I am personally taking also out some green leagues and countries.

For example, Countries and leagues with a high amount of alerts, but small profits.


Not willing to make like 200 bets in a country where my profit stands on 2-3 units. Too much job for this Yield and I will get these 2-3 units from some other models with much less work using this statistics tool.


statistics tool

 

Conclusion

After removing some leagues and countries pointed out in the previous chapter, I'm looking at different results already. 1074 alerts are perfect sample size for me. Yield is a nice 12%, and I have removed a lot of countries and leagues where I'm not able to bet on anyway. Keep in mind that this 12% Yield is yours!

No big odds and lines dropping straight after the alert sent out, no night games or very low leagues almost impossible to get on.

 


Sie können auch mögen