Week 18 is in the books! After the massive success of Week 17 (where we saw a huge ROI spike), we were hoping to secure back-to-back winning weeks. Unfortunately, the variance of the Premier League bit back slightly, and we finished the week just shy of the green line with a total net result of β¬-0.69.
While we didn't turn a profit, the week provided some fascinating data pointsβspecifically regarding the precision of our Correct Score (CS) model and the recovery of our Over/Under (O/U) strategy.
π Matchweek 18 Details
π Model: Week 18 (1u Flat)
| Model | Hitrate | Profit | ROI |
|---|---|---|---|
| 1X2 | 50% | -0.93u | -9.3% |
| O/U | 60% | -0.15u | -1.5% |
| CS | 30% | N/A | N/A |
| TOT | -1.08u |
Correct Score: The "Sniper" Strikes Again
The most remarkable stat of the week is undoubtedly the Correct Score performance. Predicting exact scores is the hardest part of football betting, yet the model nailed 3 out of 10 exact scores:
- Arsenal 2-1 Brighton β
- Nott'm Forest 1-2 Man City β
- Liverpool 2-1 Wolves β
While the CS ticket itself didn't profit (due to the specific combination of legs), hitting a 30% strike rate on exact scores suggests the model's underlying algorithms for calculating team strength are very calibrated, even if the match winner predictions (1X2) faltered.
1X2 Struggles
The standard Match Winner (1X2) market was the weak link this week. With a 50% hit rate (5/10), it is mathematically very difficult to build a profitable accumulator. The model correctly predicted strong home wins for Arsenal and Liverpool, but upset losses for teams like Chelsea hurt the bottom line.
π History per Week
| Week | Games | 1X2% | 1X2 P/L | O/U% | O/U P/L | CS% | CS P/L | Total P/L |
|---|---|---|---|---|---|---|---|---|
| 13 | 10 | 50.0% | -0.96u | 50.0% | -1.06u | 0.0% | 0.00u | -2.02u |
| 14 | 10 | 40.0% | -2.91u | 70.0% | +2.52u | 10.0% | 0.00u | -0.39u |
| 15 | 10 | 50.0% | -1.23u | 40.0% | -3.59u | 0.0% | 0.00u | -4.82u |
| 16 | 10 | 70.0% | +3.53u | 50.0% | -1.40u | 0.0% | 0.00u | +2.13u |
| 17 | 10 | 70.0% | +4.94u | 50.0% | -1.90u | 50.0% | 0.00u | +3.04u |
| 18 | 10 | 50.0% | -0.93u | 60.0% | -0.15u | 30.0% | 0.00u | -1.08u |
| TOT | 60 | 55.0% | +2.44u | 53.3% | -5.58u | 15.0% | 0.00u | -3.14u |
π« Ticket Overview (Week 18)
β 1X2 Ticket | Profit: β¬-2.20
| Match | Selection | Res | |
|---|---|---|---|
| Arsenal v Brighton | Home | Won | β |
| Nott'm Forest v Man City | Away | Won | β |
| Chelsea v Aston Villa | Home | Lost | β |
| Burnley v Everton | Away | Lost | β |
| Sunderland v Leeds | Home | Lost | β |
| Brentford v Bournemouth | Home | Won | β |
β O/U Ticket | Profit: β¬+2.61
| Match | Selection | Res | |
|---|---|---|---|
| Brentford v Bournemouth | Over | Won | β |
| Liverpool v Wolves | Over | Won | β |
| Arsenal v Brighton | Over | Won | β |
| Man United v Newcastle | Over | Lost | β |
| Chelsea v Aston Villa | Over | Won | β |
| Burnley v Everton | Under | Won | β |
O/U Strategy: Adjustments Paid Off
In recent weeks, the Over/Under predictions were struggling. We made adjustments to the selection criteria for the O/U Ticket this week, and it worked.
- The Result: The O/U ticket was a winner, yielding a profit of β¬+2.61.
- The Miss: We missed a "clean sweep" by just one goal (Man United vs Newcastle ended 1-0, missing the Over).
This winning ticket offset the losses from the other categories, nearly bringing us to break-even for the week.
β CS Ticket | Profit: β¬-1.10
| Match | Selection | Res | |
|---|---|---|---|
| Burnley v Everton | 0-1 | Lost | β |
| Chelsea v Aston Villa | 2-1 | Lost | β |
| Nott'm Forest v Man City | 1-2 | Won | β |
| Sunderland v Leeds | 2-1 | Lost | β |
π° Financial Summary per Week
| Week | 1X2 P/L | O/U P/L | CS P/L | TOTAL |
|---|---|---|---|---|
| 13 | -2.20β¬ | -2.20β¬ | -1.10β¬ | -5.50β¬ |
| 14 | -2.20β¬ | -1.18β¬ | -1.10β¬ | -4.48β¬ |
| 15 | -2.20β¬ | -2.20β¬ | -1.10β¬ | -5.50β¬ |
| 16 | -2.20β¬ | -2.20β¬ | -1.10β¬ | -5.50β¬ |
| 17 | -0.81β¬ | +2.28β¬ | +52.17β¬ | +53.64β¬ |
| 18 | -2.20β¬ | +2.61β¬ | -1.10β¬ | -0.69β¬ |
| TOT | -11.81β¬ | -2.89β¬ | +46.67β¬ | +31.97β¬ |
π° Total Weekly Net: β¬-0.69
Disclaimer: Educational experiment. 18+ Only.
We didn't get the green week we wanted, but β¬-0.69 is effectively a scratch. The data confirms that our specific score predictions are dangerous (in a good way) and that the O/U filter is improving.
Cheers,
Peter