- Nov 10, 2018
The obvious way that comes to mind that's consistent within this system is you add up the ATP score of the participants at the start of the event (or top 8 or whatever, so it doesn't just get bloated by 256 players) and use that to make some kind of multiplier. So there is no subjective judgements needed, other than those already existing within the system.How would you start determining which opponents merit a positive impact on modifiers and which lower it?
This would work better the longer a history you have so you have more accurate readings with more time to filter through. Not given it a tremendous amount of thought though.
One idea is you take the sum of the top 8 ATP players in the tourney and divide it by the top 8 ATP generally and use that as a multiplier, so any tourneys where all top 8 players are involved will be unaffected (since we assume these must be highly competitive, perhaps not exactly true but seems a decent estimation) and any where the participants are less accomplished - as measured by ATP - will be scaled down since they are deemed to be not as high level. One big problem with this is you can't do it from the start of the ratings since it doesn't work until there are somewhat sensible ratings for at least 8 players and it gives 0 puntos for tourneys where no players are rated yet which seems wrong, so maybe u can add like 0.2 to the multiplier at the end to fix both of these. Idk this was basically just the first most obvious and simplistic idea and is obviously not fine tuned, just wanted to give it as an example of how a system to judge the participants could work.
Yes this does mean that if a player with a higher ATP rating joins a tourney it makes it worth a bit more but it wouldn't be a huge impact since they are only one of 8 whose scores are being taken, so the "bonus" will still be pretty low unless there are also other top level participants (which seems to fit well with what we want to measure, if you accept that this is a reasonable thing to do - which it may or may not be).