First of all, what a fantastic weekend the last was: 20/21 winners, a new PB for us and some great interactions with those who follow and got involved.
Now..It’s easy to get carried away and start eyeing up that discounted Sergio Georgini jacket but if you’ve read past posts, you’ll know we’re all about methodical progression and it’s important to be clear – there will be losses (💲blah💲blah💲blah💲). But seriously we do look to steadily improve the way we tip and for this post I’ve done some analysis around our BetScore feature included in each match tweet.
I believe this concept of BetScore is important as it differentiates our service and attempts to give an objective view of the reasoning behind a tip compared to our peers – remember we are not affiliates and are not in the game of impulsive betting.
I’ve taken the BetScore’s tipped for each of the last 6 weeks:
Average winning tip: 61.4%
Average losing tip: 63.6%
Average tip = 63.5%
As we currently tip 50% and above we want winning tip scores to be high above the average tip and the losses to be under the average tip score as it shows greater certainty in the spread of scores we give. ie.the result of that 80% Serie A game coming in skews our winning average tip up showing we have more certainty in those games. It’s the converse effect with the losing score as the greater the gap to the overall average then it shows those riskier 50% scraped Ligue 2 games that end up not coming in are distinctly lower for a reason. Once we have established this gap consistently then we can refine the score required in order to get tipped so our overall average is more aligned to the winning average.
What’s emerging in the results so far is that we lack this gap between the average tipped and the average win/loss score and we actually have weeks when the higher BetScores do not always correlate with increased probability of a goal. The results on the 21st Jan shows the biggest difference where the average losing tip is significantly higher than the winning. Although note that the average losing tip is skewed here due to a small sample of losers and them being particularly high BetScores – 83.33% in this example. So they heavily impact the stats when they don’t come in.
Looking into some of the losses, 3 themes may be occurring:
> Both teams may lack quality to break through despite pressure.
> Sending off has occurred early in 1st half (bloating BetScore) then 10 man team parks the bus.
> Big gulf in quality between teams – Pressure in 1st half then worse team try to hang on for a draw.
> Anomaly reported in in-play stats causing extremely high BetScore.
So could these areas add some justification to why the average losing bet score is overall a higher % than the average winning BetScore?The first bullet may be the toughest to measure, as its a thin line between a goal frenzy with poor defences and poor attacks with a stubborn back line. The second could be the quickest to detect but receiving the info that there has been a sending off may be the issue there.There might be an opportunity on the 3rd point to detect ‘too much’ pressure in the first half and infer that the weaker team will try to hold out for a goalless draw in the 2nd half – this could be represented by an actual reduced BetScore (countering the assessment of high pressure). The fourth is a risk that just requires managing at the moment. We do have finite sources of in-play stats especially due to covering so many matches so we will continue to logically detect anomalies and mark them as void if we can foresee an unrealistic bet score formulating.
The fact we (albeit from a small sample) are seeing wins towards the lower bet scores, does this also mean that some characteristics of matches around and below 50% should not necessarily be disregarded?
If we were tipping games down to 45% BetScore, then the average tipped would move towards 55%, would this mean the average losses are still just above, around 57% and winners also around 55%? Or would this be the tipping point and the average winners hover around 55%? As Andy@GamblingPanel said in last post, you have the option to use this information now and bet <50% via the current matches page so would be good to hear how you get on.
What would be interesting is if the success rate of winners does remain similar between 45-50% but drops when the score is under this range then there would seem to be a balance between teams with insufficient quality to get going and those games which have such a vibrant first half (causing the bet score to bloat to 80+) but eventually peter out. This lends itself to being wary of games with a bet score >80% and perhaps more accepting of games where pressure starts lower but other conditions are right. In this sense we will continue to demonstrate our understanding and match experience with the BetScore, picking up on some of these less obvious characteristics to predict the eventuality of a goal rather than acting as a mere half time pressure gauge.
Overall our strategy is first to highlight the gap in BetScore’s between winners and losers and then ultimately move our tipping BetScore towards the average winner. Always worth noting is that this sample of data is small so we could see a different picture by season-end. A reason for this scepticism might be last weekend’s blockbuster where the higher bet scores were noticeably more successful than the lower – the trend we want to see.
We have been busy back-office the last few weeks trialing different variations in an effort to tune the BetScore for the season run-in. We are also getting our hands on more statistical factors to include in the algorithm so keep your eyes peeled for an upgrade soon.
Ones to Watch:
Bastia v Monaco – (Friday 19:45 – Ligue 1)
I like a Friday night O2W to get the ball rolling on the weekend’s action! Monaco are absolutely flying at the moment and Bastia have lost last 2 and deep in a relegation battle. Although we have such a gulf in class here I think this forces Bastia to start defensively and try to get to half time.
Huddersfield v Man City – (Saturday 15:00 – FA Cup)
Huddersfield have won their last 6 and although City seem to have turned the corner again i feel the terriers will start bright here and offer little space to Pep’s technicians. Of course the legs can’t go on forever and I’m hoping for them to tire 2nd half and City to cruise home.
Oostende v Anderlecht – (Sunday 17:00 – Belgian First Division A)
Right, time for Anderlecht to put things right after we picked them a couple of weeks back with quite ridiculous half time stats. These two are towards the top of the table and both in good form so the quality is there and I’m just hoping they wait until the 2nd half to find the back of the net.