Freezing Saddles 2015

[cvcalhoun]

@vvill 100033 wrote:

I think the mileage estimates and therefore handicaps are based on what you think you will ride during FS, not what you ride during the rest of the year.

Last year, we were asked what we had done during certain months (October/November, maybe?), rather than what we expected to do. And to me, that makes a lot of sense. If you based handicaps on what people thought they were going to do during FS, a) it would be impossible to measure whether they were telling the truth, and b) the team that won would automatically be the one in which people most wildly underestimated what they would actually do. If you use actual numbers (perhaps verified using Strava records, at least for those already in Strava), you provide an incentive for people to increase their riding, rather than an incentive for them to underestimate their future riding.

Basing the handicaps on what people did last year could work for those in the competition both years, but would still leave open the question about what to do about people who just joined this year. Those people might or might not even have records of what they did last year. And even if they did, the newbies would be at a big advantage in not having their numbers from last year artificially inflated by the competition itself.

TL;DR: Last year’s system worked fine, except perhaps for sandbagging, and if it ain’t broke, don’t fix it. Sandbagging could be avoided by basing handicaps on Strava records, at least where they exist.

[hozn]

Yes, I am on board. Changing jobs so busy right now, but the software is in a fine state to start working when we are ready. (There are many things I would like to improve, but they are not prerequisites for starting to track miles / scores etc..). The dry run system in Dec worked well to ensure everyone was signed up.

For people that have participated in previous years we can incorporate past numbers. I will give this more thought, not sure if I have any strong opinion. On one hand, it may not be fair to strictly take the handicap based on the mileage, since people push harder than they typically ride. But we also don’t want sandbagging. Last year seemed to be a bit more dynamic, but there were definitely see teams that seemingly had little hope if victory.

http://freezingsaddles.com to review the data from last year.

If anyone wants raw data to chew over, shoot me a PM.

[sjclaeys]

Have already spent more time at work than I should reading this thread. Freezing Saddles 2015 is already a success.

[vvill]

@cvcalhoun 100042 wrote:

TL;DR: Last year’s system worked fine, except perhaps for sandbagging, and if it ain’t broke, don’t fix it. Sandbagging could be avoided by basing handicaps on Strava records, at least where they exist.

Yeah I agree. The pool idea is hopefully to make it easier for people to set up (I’d hope?) because it’s self-nominated and it’s not quantitative. Since any estimate/prediction is not very precise there’s no point projecting it onto a more complex number-based team seeding system.

@hozn 100043 wrote:

http://freezingsaddles.com to review the data from last year.

Web application could not be started

[GB]

@hozn 100043 wrote:

For people that have participated in previous years we can incorporate past numbers. I will give this more thought, not sure if I have any strong opinion. On one hand, it may not be fair to strictly take the handicap based on the mileage, since people push harder than they typically ride. But we also don’t want sandbagging. Last year seemed to be a bit more dynamic, but there were definitely see teams that seemingly had little hope if victory.

I like the idea of discount points above your stated milage, and bonusing points below. I’d do a 10% bonus or discount with a +/- 10% window on your stated miles. If you think you’re going to ride 100 and you say 50 then the miles after 55 only count .9. and if you end up only riding 40 then each mile counts 1.1. This gives you a real incentive to give a good guess and to ride when your lagging behind. This should rest every week. The reason not to say 1000 miles/week and load up on the bonus points is that the teams are built to be even based on your stated milage. So although your personal score would be good your team would blame you for falling short of your goal.

[dasgeh]

Some of us have a harder time estimating miles than others. Last year, I had a number of trips in Oct-Dec and none during FS, so that was not an accurate basis. This year, my work situation is changing in a yet-uncertain way. If randomizing works (either with buckets or without), then that’s my vote.

[hozn]

@vvill 100045 wrote:

Web application could not be started

Errrr … I guess the recent server upgrade did in fact affect me I will get on that when I have some free time — prob not tonight.

[cyclingfool]

@hozn 100050 wrote:

Errrr … I guess the recent server upgrade did in fact affect me I will get on that when I have some free time — prob not tonight.

Dammit! What are we posting you for?! Oh, wait.

[ShawnoftheDread]

@dasgeh 100047 wrote:

Some of us have a harder time estimating miles than others. Last year, I had a number of trips in Oct-Dec and none during FS, so that was not an accurate basis. This year, my work situation is changing in a yet-uncertain way. If randomizing works (either with buckets or without), then that’s my vote.

You weren’t supposed to estimate. You were supposed to submit your actual miles and days ridden during the months in question. Not what you would have ridden if a, b, and c did or did not occur. It seems like last year a lot of people reverse sandbagged, stating higher miles than they actually rode because they knew they were going to try harder during the contest than they had in the fall. Just state your actual rides and we’ll all be starting from the same point.

[dasgeh]

@ShawnoftheDread 100053 wrote:

You weren’t supposed to estimate.

There was at least one proposal to estimate. That is problematic

Using October – November data is also problematic because some of us are going to have more month to month variance than others (eg those of us changing jobs that halve our commute)

[ShawnoftheDread]

@dasgeh 100054 wrote:

There was at least one proposal to estimate. That is problematic

Using October – November data is also problematic because some of us are going to have more month to month variance than others (eg those of us changing jobs that halve our commute)

Yes, everyone has an issue with something. Some of us hate the cold so we’re at a disadvantage in a contest that emphasizes winter riding. Riders from the Midwest should be handicapped differently than those from California.

[peterw_diy]

Teams of 10, randomly selected. I love the idea that random selection means no worries about sandbagging (last year I stressed a bit because I inadvertently sandbagged by underestimating some normal route lengths), and less trouble for any late team assignments.

Also seems a relatively easy programming enhancement would be team size handicapping. Somebody drops out right before the start? That team gets an 11% mileage bonus. Some team has 11 riders? 9% mileage penalty.

OK, so, jrenault, hozn, let’s discuss what PRNG to use… Dual_EC_DRBG?

[rcannon100]

Last year there was a non-competitive team (a team participating in FS and pointless prizes, but not competing in the mileage / points competition.) My assumption is that option will be offered again as well.

[dasgeh]

@ShawnoftheDread 100057 wrote:

Yes, everyone has an issue with something. Some of us hate the cold so we’re at a disadvantage in a contest that emphasizes winter riding. Riders from the Midwest should be handicapped differently than those from California.

It depends on what the goal of handicapping is. I don’t think anyone is saying handicapping should control for how “hard” it is on a particular person. My understanding is that handicapping is to balance teams. My point is that neither an estimate model nor a past performance model will work to achieve that goal because some people will have circumstances that are unique.

I’m hoping randomization will.

[DismalScientist]

@ShawnoftheDread 100057 wrote:

Riders from the Midwest should be handicapped differently than those from California.

Hey, I’m already handicapped by my personality.

[Author]

Posts