How ratings work
Every sport has its own rating, all on one scale. 8.00 is the middle of your group, and a point is one clear step up: winning about 64% of the games at tennis, one gym grade at climbing, a 5 km that's 12% quicker at running. UMR is your two best sports combined.
Nothing is stored. Every rating is worked out again from the raw logs each time a page loads, so when a rule changes, everyone's history is re-scored the same way.
The scale
- Ratings are relative to your group. An 8.00 is the middle of this group, not of the world. A 9 here isn't a 9 in another group.
- Tennis starts everyone at 8.00, and results pull people apart.
- Climbing and running have units of their own (grades, minutes), so each is shifted by one number that puts the group's average at 8.00. That shift follows the group until three mates have enough logged (five sends each for climbing, one run each for running). Then it locks, so someone new joining can't move everyone else.
Why 8, and why one scale? Eight is about mid-table on UTR, the tennis rating this started as a joke about, and it leaves room underneath for mates who are just starting. One scale is what lets UMR combine sports: a point has to mean about the same amount of better in each of them.
Tennis
Underneath, this is Glicko, the rating system chess sites use, adjusted for a few mates who play each other a handful of times a year. It isn't UTR's formula, which is private, but it borrows UTR's main idea: games count, not just wins.
- What was expected. Before a match, the two ratings say how it should go. A mate one point above you should take about 64% of the games and about 85% of the matches.
- What happened. Your result is 75% the share of games you won and 25% whether you won the match. Beat what was expected and you go up; fall short and you go down.
- How much was played. A match of 20 games or more counts in full, and shorter ones in proportion: one set at 6-4 is 50%, a Fast4 set at 4-2 is 30%, a lone tiebreak 15%.
- Doubles counts 50%. A pair is as strong as the average of its two players, and both partners move.
- The winner never loses rating, however narrow the win.
- New players move fast. The system is unsure about someone new, so their first few matches move them a lot and later ones less. A rating wears an asterisk until three matches.
- Rust. After 60 days without a match, your form drops 0.25 a month, up to 1.00. It comes straight back the day you play: rust is taken off what's shown, not off the rating underneath.
What one match does
Between two evenly matched mates who have each played a dozen matches. Newer players move more.
| Result | Winner | Loser |
|---|---|---|
| 6-0 6-0 | +0.19 | −0.19 |
| 6-3 6-3 | +0.14 | −0.14 |
| 6-4 6-4 | +0.12 | −0.12 |
| 7-6 7-6 | +0.09 | −0.09 |
| 0-6 7-6 7-6 | +0.05 | −0.05 |
| One set, 6-4 | +0.06 | −0.06 |
| Lone tiebreak, 7-5 | +0.02 | −0.02 |
| Doubles, 6-3 6-3 (each player) | +0.07 | −0.07 |
Why games, not just wins? Friends don't play each other often. Win or lose is one fact per match; the games say how close it was, so ratings settle in far fewer matches.
Why does winning still count for 25%? On games alone you could win 0-6 7-6 7-6 and go down, because you lost more games than you won. This share is enough that between equals the winner always goes up and the loser always goes down, while 6-0 6-0 is still worth far more than 7-6 7-6.
Why is there nowhere to enter a tiebreak score? A set that went to a tiebreak goes in as 7-6. The tiebreak's own score would move a rating by less than it rounds to.
Climbing
- Each send is worth its grade at Sydney Indoor Climbing Gym, St Peters (1 to 10). A flash adds 0.5.
- Real rock is worth more. An outdoor V-grade plus 3 is the home grade it feels like, then 2 more for being outdoors: V2 feels like a home 5 and is worth 7.
- Your rating is the average of your best five different problems, at what each is worth today. Sends worth 7, 6.5, 6, 6, 5.5 make a 6.2.
- Sends fade. A gym send keeps its full value for 30 days, then loses 0.5 a month. An outdoor send holds for 180 days, then loses 0.25 a month. A home 6 from 90 days ago is worth 5 today.
- Until you have five, each empty slot costs 0.25 and the rating wears an asterisk.
- Only sends and flashes are logged, never attempts, so there's nothing to lose by logging and nothing to gain by hiding a bad day.
Why five? One send can be a soft grade, or a style that happens to suit you. Five is about one good session, so a rating fills in a night and a single fluke can't carry it. The big climbing logbook 8a.nu ranks on a climber's ten best in a year; gym walls are reset every few weeks, so ten would rarely fill.
Why fade at all? The rating is meant to say how you're climbing now. A gym problem is stripped within weeks, so an old send says less and less about you. Rock stays put and outdoor days are rarer, so those hold six times as long.
Why the outdoor bonus? No coloured holds, fewer pads, and you had to get there. Real rock should weigh more and last longer.
Running
- Log a distance and a time. Any run from 3 km to 50 km counts.
- Every run becomes a 5 km time, using Riegel's formula, the one behind most race-time predictors: time × (5 ÷ distance) to the power 1.06. A 10 km in 50:00 is worth 23:59 for 5 km.
- Every 12% quicker is a point. 25:00 is one point above 28:00, and 22:19 is one above 25:00.
- Your rating is the single run worth the most today. A slower run never lowers it, so log the easy ones too.
- Runs fade. A run keeps its full value for 60 days, then loses 0.25 a month, so a rating drifts down unless you show it again.
| 5 km time | Rating, if 25:00 were 8.00 |
|---|---|
| 18:00 | 10.90 |
| 20:00 | 9.97 |
| 22:30 | 8.93 |
| 25:00 | 8.00 |
| 28:00 | 7.00 |
| 30:00 | 6.39 |
| 35:00 | 5.03 |
| 40:00 | 3.85 |
Why your best run, not your average? Most runs are meant to be easy. A rating built on the average would punish anyone who logs their recovery jogs, and people would stop logging.
Why percentages, not minutes? Taking a minute off 30:00 is far easier than taking one off 18:00. Equal percentages are about equally hard wherever you start.
Why 12%? It spreads a group of social runners about as widely as tennis and climbing spread theirs: a 20:00 runner and a 30:00 runner are 3.6 points apart, much like a strong and a weak club player, or three or four gym grades. Otherwise running would count for too much or too little in UMR.
UMR
The Universal Mates Rating: universal across exactly one friend group.
- It's your two best sports, out of the ones ticked in Settings, combined as a geometric mean (multiply them, then take the square root).
- Being lopsided costs you. 8.00 and 8.00 make 8.00. 9.00 and 7.00 make 7.94. 10.00 and 6.00 make 7.75.
- One sport isn't enough. You're on that sport's board, and shown as "one sport short" on the UMR board until you log a second.
- Another sport can't hurt. Only your best two count, so trying something new never lowers your UMR.
- It wears an asterisk while either sport behind it does.
Why your best two, not every sport? Not everyone does everything. Counting every sport would leave the mate who doesn't run with no UMR at all, and would make adding a sport a risk for everyone already in the group.
Why a geometric mean? UMR is meant to reward the all-rounder. A plain average treats 10 and 6 the same as 8 and 8. Multiplying doesn't.
Peaks, arrows and asterisks
- Form is your rating today, after fading and rust.
- Peak is the highest your form has ever been, with the date. It never fades.
- Peak UMR is the highest the combined number reached on a single day. You can't stitch one year's tennis to another year's climbing.
- Arrows show the change over the last 30 days.
- An asterisk means provisional: fewer than three matches, or fewer than five sends.
Who checks the scores?
Nobody has to. Anyone in the group can log anything and it counts straight away; a photo or video is optional. If a match score is wrong, anyone else who played can tap "That's not right". It comes out of the ratings until it's sorted, and can be put back. Whoever logged something can remove it.
Rough edges
Known, and either left in on purpose or not fixed yet.
- A tennis favourite who wins by less than expected gets +0.00 while the loser gains a little. That's the rule that a winner never loses rating, and over many months it lets tennis ratings creep up.
- A big underdog can still gain from a close loss.
- Upsets swing hard: a favourite who loses in three sets can drop about 0.4.
- Until you've logged five climbs, logging one well below your others can lower your rating, because it pulls the average down by more than the empty slot was costing. It stops once five are in.
- The outdoor conversion (V-grade plus 3) is a guess. An admin can change it in Settings.
- Running assumes a flat course and treats everyone alike: nothing for hills, trails, age or sex. Runs under 3 km don't count because the conversion flatters sprinters.
- Ratings from different groups can't be compared.