I’ve tentatively re-initiated my predictive previews to see whether there’s any mileage in making them a regular feature, starting with this weekend’s Championship games but with the intention of expanding to cover all Football League divisions. I wanted to make them a bit more visually engaging than the standard blog format would allow, so I’ve created a PDF template, the style of which will be familiar to readers of my milestone reports:

Each match has a separate page which shows up to 3 key head-to-head stats, incorporating both a bubble and a mini-graph showing how that team stacks up against the rest of the division, and an overall summary prediction at the bottom. This is definitely NOT gambling advice – if you saw how laughably inaccurate my predictions were the last time I tried this, you’ll understand why I won’t putting any money on these.

This is very much a first pass at doing this, so feedback is definitely encouraged.

For those who are interested, the predicted scores are generated by:

• Taking the shots per game and goal data which powers the attack and defence visualisations;
• Weighting the data in favour of more recent performances;
• Applying adjustments to reflect identified strengths and weaknesses, with the main ones being blown out in the preview.

For example, if a team create a large proportion of their goals down one side of the pitch and are playing a side who concede a lot from that area, this would raise their goal expectation, while a team which tends to score a lot of headers but are up against a side that concede relatively few would see their goal expectation reduced.

I’m still calibrating the model and I’m not sure it’ll ever be a reliably accurate predictor as it doesn’t have enough ‘proper’ statistics in it, but it still adds a bit of texture to the previews.