Premier A: Redland B v University of Bristol Chiefs (Tue 08 Mar 2011)

Match played between Frank Ellis (home) and Nikhil Shah (away).
Match won by Nikhil Shah. Result: 7-11,4-11,10-12.

Starting level for Frank Ellis: 5,309, level confidence: 75%. Set manually.
Starting level for Nikhil Shah: 2,610, level confidence: 58%.
Frank Ellis to win as he is currently playing 103% better than Nikhil Shah.

Nikhil Shah won all of the games and 62% of the points.
This games result would be expected if he was better by around 55% or more.
This points result would be expected if he was better by around 62% (PAR scoring).
These are weighted and combined to calculate that Nikhil Shah played 62% better than Frank Ellis in this match. An upset!

Assuming that any level changes are shared between both players, for this result it looks like Nikhil Shah actually played at a level of 4,736 and Frank Ellis at a level of 2,925. Without any damping, both players would need to be adjusted by 81% to match this result.

Allowing for the difference in level between the players, the adjustments have been reduced to 52% and 52% respectively.

Factoring in the relative levels of confidence which allows players with low confidence in their levels to change more quickly, the adjustment for Nikhil Shah changes to +52% and Frank Ellis changes to -37%.

After applying standard match damping, the adjustment for Nikhil Shah becomes +17.1% and for Frank Ellis becomes -10.4%.

Given Frank Ellis's level and the type of match played, an additional damping of 1.3% has been applied to his level change. Looks like he wasn't taking the match too seriously...

Apply match/event weighting of 75% for 'Mixed Spring 2011' so the adjustment for Nikhil Shah is +13% and for Frank Ellis is -7.5%.

Apply limits to the amount of change for a single match which are based on player level, level confidence and time since last match so that Nikhil Shah is limited to +10% and Frank Ellis is limited to -5% level change. In general a player's level won't go up by more than 10% or drop more than 5% if they've played in the last 7 days but those limits are relaxed if their previous match was further back.

Increase level confidence due to one more match played. Nikhil Shah: 76%, Frank Ellis: 86%. Reduce level confidence based on how unexpected the result is. Nikhil Shah: 42%, Frank Ellis: 48%.

A final adjustment of +0.5% has been made to both players as part of the automatic calibration that is performed after each match. All players in this pool will have been adjusted equally in order to remain equivalent to other player pools.

Final level for Frank Ellis: 5,112, level confidence: 48%.
Final level for Nikhil Shah: 2,867, level confidence: 42%.

Notes

• This calculation is done in two main parts; first work out the adjustment needed to match the results and then apply damping. This means that levels should always be 'about right' but the time taken to get there or the volatility is dependent on the damping.
• A level also has 'level confidence' which drops if players haven't played for a long time or have had enexpected results. As low confidence levels adjust more quickly than high confidence levels, it allows these players to find their level more quickly without impacting their opponent's level too much.
• Point scores are used as well as game scores for accuracy - particularly important for 3-0 results - though we can work with game scores only too, albeit with more damping.
• Mismatched players are allowed for - you don't have to hammer your opponent. See explanation above if this applies to this match.
• The section on damping is where we still have some options. We have recently made a change to damp league matches more than tournament matches and box matches even more than that. This gives added weight to the more important matches.
• There are occasional, very small adjustments made to all players to keep the averages constant which are not covered here.
• You don't get a bonus just for winning - if you want to go up you have to play better than expected against your opponent.
• We have spent more than 5 years fine tuning the level calculations based on tens of thousands of match results and a great deal of feedback from players, team captains and coaches. It's the most usable and accurate ranking system there is in any sport, let alone squash.
• For a more complete explanation of how levels are calculated (on which this system is based) see the help file here.
• If you have thoughts/opinions on the above or any feedback on the way levels are calculated or updated, please contact us. We welcome all feedback, although we are keen squash players ourselves and would pefer to be on-court than in front of a screen so please be patient and please do try to see if your question has already been answered on the help page. We are unable to answer questions about hard anyone played in their match - we only get to see the results - and if your level didn't increase as expected please make sure you've looked at the above explanation before contacting us. If you want to go up the levels, train harder, listen to your coach and win more points. Or just be incredibly talented!