The future of ccRating

[timbaeyens]

I have another remark towards Cc rating: in the top 20, there are pilots who haven't flown in Cc-rated flight since 2017 and even since 2012; since Condor 1 thus.

That feels hugely wrong.

[witor]

They do not appear on the list of active pilots. Changing the numbering would be just a cosmetic change, but I agree it would be nice to have.

Why does it feel wrong? They have earned their rating by winning races. How can you tell their their skill is lower now?

Would it be OK to delete a pilot and then when he comes back give him a low rating?

In that case, everyone's rating will suffer because when he comes back and starts winning again, because of his low rating field strength of the race will be lower so winning would not give you that many points despite the fact that you just won against very strong (and highly underrated) pilot who in near future will most likely come back to 3500 level.

I don't know about you, but I do not want to compete against some of the top pilots if they have a rating of 1000 because they were inactive and recently reactivated... Winning would be very difficult and losing a race would cost a million rating points.

I said this already numerous times: The only way to lose or gain rating points must be through winning or losing against other pilots (players).

[velaskez]

FIDE moves players to an inactive list, if they do not play ranked games for a year. They do however keep they're rating and are listerd on the All Players ranking forever. That's why Kasparov is 2nd in the world even though he hasn't played since 1999. Maybe that is something we should consider? We kind of have it already with the All Pilot's list and the yearly cumulative score.

[timbaeyens]

They do not appear on the list of active pilots. Changing the numbering would be just a cosmetic change, but I agree it would be nice to have.

Agreed.

Why does it feel wrong? They have earned their rating by winning races. How can you tell their their skill is lower now?

Well, Condor 1 cannot be compared to C2, and definitely not to C3. I have been mostly flying on the conservative side since C1, and with every evolution, I get better scores. We saw others struggling with C3. The simulation has gotten better, but very different.

Would it be OK to delete a pilot and then when he comes back give him a low rating?

In that case, everyone's rating will suffer because when he comes back and starts winning again, because of his low rating field strength of the race will be lower so winning would not give you that many points despite the fact that you just won against very strong (and highly underrated) pilot who in near future will most likely come back to 3500 level.

I don't know about you, but I do not want to compete against some of the top pilots if they have a rating of 1000 because they were inactive and recently reactivated... Winning would be very difficult and losing a race would cost a million rating points.

I said this already numerous times: The only way to lose or gain rating points must be through winning or losing against other pilots (players).

I understand your reasoning, but imagine those pilots reached places 1,2 and 3 in 2012, and never fly again. Would you, Arne or any current top pilot ever be able to get to the podium? (Honest question, I don't know how CC rating is calculated).

[witor]

with every evolution, I get better scores.

Because you are getting better as you are gaining experience. People who were good in C1 are still good in C3.

As we both agree inactive pilots are not on the main list. I think this is a good compromise. I already said that I would also prefer the numbering not to include the inactive pilots if they are not displayed so someone in 20th position amongst active pilots could quickly see where he is at without manually counting. We are on the same page on that. Unfortunately, this is not up to me. I can't remember if there was a reason Thierry made it this way on CC.

[wickid]

If there is something that needs to be fixed it is definitely this:

Yesterdays CWC, Frank (O9) was number 3 in the ranking, and he crashed at the start and got his rating reduced by 143 points. I was in rank 8 before the race. I safely landed out after 110 km and I get 142 penalty points. That is not correct in my eyes. Sure, I landed out, so I have to lose rating points. But why only 1 less than somebody that crashed???? Why don't I get any credit for 110 km? Surely landing out should be way less harshly penalised than crashing.

Maybe it is due to CWC not marking "crashed" with 0 points but giving distance points. That is also not correct in my eyes.

[timbaeyens]

with every evolution, I get better scores.

Because you are getting better as you are gaining experience. People who were good in C1 are still good in C3.

Partly only. I have been flying RL since 1984; the more Condor evolved to a RL experience, the more my RL flying style (which I have kept through C1, C2, C3 for analogy purposes) suits the simulator.

[ryanwoodie]

I already said that I would also prefer the numbering not to include inactive pilots if they are not displayed, so someone in 20th position among active pilots could quickly see where they stand without manually counting. We're on the same page on that. Unfortunately, this isn't up to me. I can't remember if there was a reason Thierry made it this way on CC.

I agree. It would be a very small change (to exclude inactive pilots from the numbering), but in terms of what it communicates and how it feels—how it fuels pride in accomplishment—it makes a big difference in the meta-game aspect of it. If I am ranked #7 because there are two inactive pilots ahead of me, do I consider myself #5 or #7? Would "pretending" it’s #5 feel artificial?

How can you tell their skill is lower now?

The only way to lose or gain rating points must be through winning or losing against other pilots (players).

Perhaps I’m reading this wrong, but I believe the goal of this conversation is to explore ways to make CCrating more representative of actual skill rather than artificial factors (e.g., BSOD).

There’s a strong argument that not applying rating decay actually skews the data artificially, which is why decay exists in many rating systems. Hear me out:

- In sports (and elsewhere), there’s a principle of "techno-physiological evolution." Take the best NBA/NHL/NFL team from the 1980s—would they be on par with any 2025 team? Nostalgically, we might want to think so, but in reality, it's doubtful. The level of competition in any activity tends to improve over time, especially in a relatively new sport. This is even more pronounced when the sport itself undergoes significant changes (e.g., C1, C2, C3) in addition to technological advancements (eg. flight computers in 2012 vs 2025).

- Skills naturally stagnate when they aren’t used. For example, the Stat Check Elo system introduced a 20% decay over 13 weeks of inactivity because it more accurately reflected player performance upon return.

Is it truly realistic to assume that a ~3500-level pilot from 2012 could return to an evolved sport, with no skill atrophy, no changes in the skills required, and no overall improvement in the competitive field? Maybe a purely cerebral game like chess has less atrophy, but Condor is not quite chess.

One possible approach that makes sense is the Universal Rating System (URS) used in chess. It recalculates ratings for inactive players by combining their past results with their performance upon returning to competition. Instead of directly penalizing inactive players, it dynamically adjusts ratings based on weighted performance, where newer games carry more weight and older results gradually decay.

Code: Select all

  
The Universal Rating System (URS) for chess employs a dynamic approach to handle player ratings, including those of inactive players:  

1. Continuous Recalculation: URS constantly updates ratings for all players using data from the past 72 months, regardless of activity status.  

2. Weighted Performance: Recent games have a greater impact on a player's rating than older ones.  

3. Exponential Decay: The influence of older games diminishes over time, following an exponential decay pattern.  

4. No Explicit Inactivity Penalty: Unlike systems that directly reduce ratings for inactivity, URS allows inactive players' ratings to naturally adjust relative to active players through its weighting mechanism.  

5. Seamless Reactivation: When an inactive player resumes competition, their new games are immediately incorporated into the rating calculation alongside their historical data, effectively updating their rating without a separate adjustment process.  

This method allows URS to maintain an up-to-date representation of a player's strength, even during periods of inactivity, without resorting to arbitrary decay rules or manual adjustments when players return to active play.  

Beyond just statistical accuracy, we also need to consider how CCrating functions as a meta-game and catalyst within the community. Does it make the sport more fun? Does it give us something to strive for? Does it encourage active participation?

Even if the above arguments for decay don’t hold, and keeping inactive players' ratings static is technically more "correct," is it the best system if it discourages participation? Could it lead to “winning by not losing (ie. not playing)”? Given that our biggest challenge as a community is its small size, we should design the meta-game to encourage engagement and participation.

[witor]

Maybe it is due to CWC not marking "crashed" with 0 points but giving distance points. That is also not correct in my eyes.

ccRating does not care if you landed or crashed. The only thing that matters is your position on the results list. It is up to the organisers to decide if they want to give distance points for crashing or treat the crash as zero points. I am not sure which way is better. Should a pilot who crashed 100m before the finish line lose with the pilot who safely outlanded after flying 10km of the 250km task?? If you try to make a risky pass land out on the top of a very steep and narrow mountain pass and another pilot's landing gets detected as a crash by condor when they tried to safely outland on a nice big field in the middle of the valley, who should win? I would be tempted to say that if you don't make it to the finish line, whoever covered a greater task distance should win. But I can see that others might feel differently. In the real world if you crash your glider (or damage it while outlanding) are you scored with zero points?

[witor]

Perhaps I’m reading this wrong, but I believe the goal of this conversation is to explore ways to make CCrating more representative of actual skill rather than artificial factors (e.g., BSOD).

Yes that was the initial goal.

There’s a strong argument that not applying rating decay actually skews the data artificially, which is why decay exists in many rating systems.

I am not sure that there actually is s strong argument. Decay might exist in some systems but the most respected ones like elo, glicko, glicko 2, trueskill on which many other were created do not have any decay mechanism. Decay just artificially decreases the rating. I am a strong believer in an idea that rating should only change as a result of direct competition between competitors. Just because someone does not fly doesn't mean they are not good anymore. After a long inactivity, SOL came back and became FAI vice world champion leaving many active pilots behind...

Glicko 2 has an uncertainty parameter. If you are inactive rating changes can be bigger next time you participate. But you still need to win or lose for your actual rating to change. I am not familiar with URS (where is it used? I never heard of it) but from what you posted there is several things I don't like about it. In particular, the decay of rating goes against my main principle that only direct competition should change your rating. You just don't know how good people are if you don't test them.

Please keep in mind that it is not easy to come up with a good rating system for condor. Note that quite often we have races with many participants. If you win against another 100 pilots some systems would unrealistically skyrocket your rating, which is not great. That's why some of the well-respected rating systems are not the best for Condor. ccRating is similar to iRating. Iracing is a kind of similar game to condor. you could be the best driver and get killed on turn one of the race and lose with 100 noobs. Because of the similarities, I think what we have is pretty good. I don't think we need a revolution.

I agree. It would be a very small change (to exclude inactive pilots from the numbering), but in terms of what it communicates and how it feels—how it fuels pride in accomplishment—it makes a big difference in the meta-game aspect of it. If I am ranked #7 because there are two inactive pilots ahead of me, do I consider myself #5 or #7? Would "pretending" it’s #5 feel artificial?

I already said multiple times that I agree that it would be better if the main numbering did not include inactive pilots. Maybe Thierry will change it one day...

[arneh]

ccRating does not care if you landed or crashed. The only thing that matters is your position on the results list.

But Wiek was scored ahead of Frank. A bad result in 65th place but still better than Frank's 76th. And his race points were certainly better at 37.5 against 0. Yet their rating dropped by almost exactly the same number (and Frank was ranked ahead before the race, so that's not why):

Screenshot from 2025-02-17 22-30-52.jpg (22.91 KiB) Viewed 551 times

[wickid]

ccRating does not care if you landed or crashed. The only thing that matters is your position on the results list.

This is not true according to the explanation:

When the race is over, the ranking is compared to the probabilities and the greater the difference between the expected and the realised, the greater the impact on the change in the pilot's rating. For example, a pilot with a low rating who places ahead of a strong pilot will get a large increase in rating. Similarly, a good pilot who is ranked last (landed out) will see his rating strongly sanctioned. This penalty is even higher if this bad position is due to a crash but is partially mitigated if there are a significant number of other crashes. Thus, this rating encourages safe flying rather than reckless risk-taking.

https://www.condor.club/generic/261/wha ... ing__.html

So I think something is not right! It can't be that I land safely and am a lot of positions higher than Frank is, but loose only 1 point less. Frank is ranked higher than me! The calculation is not correct I think.

And as it says, CC-rating should encourage safe flying. So a crash should have an extra penalty compared to a safe outlanding.

[Vertigo]

Beyond just statistical accuracy, we also need to consider how CCrating functions as a meta-game and catalyst within the community. Does it make the sport more fun? Does it give us something to strive for? Does it encourage active participation?

If ccrating wants to measure skill, then perhaps instead of trying to change it, we should look at making a different rating that represents pilots performance in competition/races over the past year or per calendar year.

iRating measuring skill is useful for racing, because you cant race with 200 people, I imagine its really annoying and unfair for a top racer to be hindered by field of noobs he has to overtake, and so the rating gives you a way to select or group drivers of similar skill. Same is true for chess, or any one on one game, it wouldnt make sense to let me compete against grand masters, you need some way to select qualified players. Its not something we need in condor; aside from maybe a few select events, like live streamed FAI races where it may make sense to group the best X players, we can have hundreds of pilots in a single race spread over different servers, it does not matter what their skill is.

[Jannes]

Just because someone does not fly doesn't mean they are not good anymore.

That's true and I see your point, but some of these ratings have been gained under very different circumstances. Imo Ryan is right saying the changes of the environment are completly ignored by the current system. But that's not necessarily a problem.

It depends on what the rating is meant to be. Currently it's an overall "Condor rating", not a "Condor 3 rating" nor a "Active Pilots rating". So as long this is what ccRating aims for, any form of rating decay would be against this.

So the remaining question is, if there is common ground for this. Alternatives would be rating decay to make it an "Active Pilots rating" or an entirely new rating based off of Condor3 alone. I guess that is not what most people would what.

I am not that familiar with iRacing. But afaik it has been around for quite some time. I guess it has iterated trough major versions like Condor did. So the environment is also not static. I think keeping that as a template for ccRating is generally a good idea.

[ryanwoodie]

I already said multiple times that I agree that it would be better if the main numbering did not include inactive pilots.

Yes, of course, I was agreeing and emphasizing the importance of changing this! I guess rhetorical questions sound like arguing but that is not the intented tone!

I am not sure that there actually is s strong argument.

I thought it was decent! IMO, weighting for recency makes results more relevant, more accurate, and most importantly makes CCrating less stagnant, and more dynamic and engaging. For the same reason we are discussing dropping scores (which is basically the same thing: changing ratings outside of competition), there are other concerns to balance, and maybe it is worth considering if having CCrating encouraging participation is a worthwhile.

Do you think static CCrating, as is, can be a possible deterrent for top (or #1 in your case) pilots to engage more freely in competition? Do you have to be more cautious or careful about what races to bother with? Wouldn't it be more exciting for #1, #2 and #3, to have a surging contender try to challenge for the podium spots vs hiding behind what happened ages ago in a different game in 2012 (this is rhetorical, I'm not at all saying YOU are doing this, I know you have great competitive spirit) ? Or have it encourage one to dust off the joystick and defend if they've been too long inactive? This is important, as having the best pilots join is maybe the greatest draw for other pilots to want to participate. So I think some weighting can be a win-win all-round.

I don't think we need a revolution.

I don't know if a weighting scores on recency would be a massive upheaval with the core algorithm, but maybe it is! If it was being considered seriously (I won't get my hopes up), it could be sanity-checked on the existing data.

but if it is a very small number (1/20, 1/25?)

Simple checkbox 'exclude this race from ccRating' would do.

Back to the low-hanging fruit solutions. 1/20 or 1/25 feels about right to me. But I'd suggest a drop-scoring (eg. every 20 races, worst race discarded) that is applied automatically versus user-directed as won't this be even worse of a situation of rating being affected by totally arbitrary non-competition forces, favouring those who are better at remembering to check the box?

I am not familiar with URS (where is it used? I never heard of it)

Here is more info from ChatGPT:
Overview of the URS

The Universal Rating System (URS) was developed to offer a comprehensive and current measure of chess strength by combining results from classical, rapid, and blitz games into one unified rating. This system continuously recalculates ratings by reassessing every game in a player's history, ensuring that the most recent performances have a greater impact.

How Recency Weighting Works in URS

• Exponential Decay of Older Results: URS applies an exponential decay function so that more recent games carry significantly more weight than older ones. This means that a game played recently influences the rating much more than a game played several years ago.
• Iterative Recalibration: As new results come in, URS not only factors these in but also re-evaluates past game outcomes in the context of current performance.

Why Recency Weighting Increases Accuracy

• Reflecting Current Form: Because players’ abilities can change due to factors such as training, injuries, or evolving strategies, emphasizing recent games ensures that the rating mirrors a player’s current strength rather than an outdated historical average.
  Citation: CHESS.STACKEXCHANGE.COM
• Enhanced Predictive Power: A rating system that prioritizes recent performances is better at forecasting near-future outcomes—a crucial factor in tournaments and competitive matchups.
  Citation: UNIVERSALRATING.COM
• Adaptive to Changing Conditions: The recency bias enables URS to quickly adjust to shifts in a player’s performance or changes in the competitive landscape.
• Constant Opponent Reassessment: URS continually updates the ratings of opponents as well, ensuring that the context of each game is recalculated based on current strengths.

Conclusion

By applying recency weighting through an exponential decay mechanism, the URS minimizes the impact of outdated results and focuses on current performance. This approach results in:

• More accurate, up-to-date ratings that truly reflect a player’s present form.
• Improved forecasting of future game outcomes.
• Enhanced adaptability in a dynamic competitive environment.

In short, the recency weighting feature in the URS is a fundamental design element that boosts its predictive accuracy and overall reliability.
Citation: EN.WIKIPEDIA.ORG