[davopnz]

So, I've always thought you should adjust the input into a model for results that are affected by things such as weather, freak occurrences, injuries, etc. The problem is, I don't really know how much you should adjust.

Some ideas I have:

1. Adjust the final score for each team up to the league average
2. Adjust the final score for each team up to the average of both teams
3. In football, add a touchdown to each teams score

Of course, there's no guarantee the game would've reached the league average had conditions been good. Low scoring games can obviously occur when conditions are perfect. One compromise might be to adjust the score to 80% of league average or something along those lines.

Also FWIW, I got this idea from an article on Tony Bloom who runs "Starlizard". Apparently they adjust soccer results for things like missed penalties, shots that hit the crossbar, etc. For example: if the score was 0-0 but one team missed a penalty they might put the result in as 0.8-0 to the team that missed.

Any thoughts?

Cheers
Enable GingerCannot connect to Ginger Check your internet connection
or reload the browserDisable in this text fieldRephraseRephrase current sentenceCEdit in Ginger×

[Waterstpub87]

Depends what type of conditions you are looking at:

For like binary things (raining or not, ect) or like a list of things (clear, raining, snowing) ect or groupings (temp under 50, temp between 50 to 70, over 70)

Get all the games and weather conditions. Put them into excel. You can use a pivot table to figure out averages for each category. Compare the to the average of all the game. You now have an adjustment for everything.

If its like a continuum (how does mph wind effect home runs), do the same but run a regression instead.