Rules and scoring thread

[Kitty]

Thanks for the info. As a data wonk, I understand how that would mess with things. If I run into issues, I’ll hit you up Judd for some gpx-fu.

As for the rule itself, Somehow I missed it last year? The sticky straight-up reads: “Manual entries will also not be accepted. This is a change from previous years.” Leading me to believe it was new this year. Without that clause it wouldn’t have seemed like a new thing. (A “the sin is in semantics” situation)

[Steve O]

The spread between the #1 team and the #22 team is currently 65%. I just ran 10 randomized sets of teams. The average (and median) spread last to first was 74%, so only barely worse than what we have. 2 out of the 10 randomized teams were actually better than 65%.

This thread included a well supported discussion of ideas that might reduce that spread, which would tighten the competition and potentially make it more fun. One idea that had a good deal of traction was to count miles 3, 4, 5 more than miles 63, 74 and 85. A couple of those ideas are copied below.

We left this thread with the suggestion that some alternative scoring systems be beta tested this year, so we would have 9 months to argue about them. I’d be curious to see how the below ideas might tighten the scores among the teams. Is there a chance that these could be test-driven? Perhaps simply add columns to the current DRASS leaderboard rather than create entirely new ones.

@Bob James 169087 wrote:

Below is an example, of accelerating returns/points. At first it appears no difference in point spread, but if that century rider skips a day of riding and a daily rider does 10 miles, they would gain 65 points on the sporadic century rider (and more importantly team points).

[ATTACH=CONFIG]15807[/ATTACH]

If someone rides at least 10 miles a day, there is no loss of points, i.e. no penalty for riding more, but was dramatically encouraged to ride daily by the scoring system. It would be very difficult to voluntarily forfeit up to 55 bonus points just because it was sub-freezing temps and/or wet conditions. It’s not really a change to the current scoring system (1 mile is still 11 points), but an extension of it, with the same goal of encouraging daily riding.

@hozn 169081 wrote:

If someone wants to propose a diminishing mileage formula that seems reasonable, I’m generally in favor of the idea. I was trying to see if I could find other similar scoring systems that had been used elsewhere, but hadn’t turned up anything in brief research.

@LhasaCM 169084 wrote:

As “food for thought” – here’s a quick example of a “tiered” system of diminishing returns for the number of miles ridden in a given day, separate from the bonus points idea. This is setup so that after every 10 mile increment, the value of the next incremental mile is reduced by, for example, 10%. In other words, 1 point per mile up to 10, 0.9 points per mile for anything between 10 and 20, 0.81 points per mile for anything between 20 and 30, etc. The Excel formula to calculate that is: 10*((1-0.9^ROUNDDOWN(MILES/10,0))/(1-0.9))+MOD(MILES,10)*(0.9^ROUNDDOWN(MILES/10,0)) That decreases a little bit faster than the previous example, with the 101st mile being worth 0.35 points. You can make it diminish faster or slower easily enough by changing the 0.9 figure either lower or higher. The table below shows what happens every 10 miles for 0.95, 0.9, 0.8, and 0.7.

[ATTACH=CONFIG]15806[/ATTACH]

One could mix and match so that you have a similar tiered idea, but with accelerating discounts (i.e., maybe you start with a 0.9 factor but make it 0.8 after 40 miles and 0.7 after 80 miles or whatever). Again – it all depends on what you want the results to look like.

[Steve O]

The other idea I’m wondering about is the “rides with” concept that was also suggested.
Would it be hard to capture the number of unique other FS participants each rider has ridden with?

[hozn]

Unfortunately “ride with” data has suffered Strava privacy api changes, so I don’t think we can actually use this data anymore.

I’d love to play with scoring, but as expected my work has kicked into overdrive until June, so I don’t anticipate much time to play with this. Would have hoped to outsource this, but so far the number of FS web volunteers hasn’t translated into any help on the backend.

[jrenaut]

I’m interested in this. In part because I’ve never been on a team with any shot at all of winning, and in part because i like mucking around with numbers. I want every year’s FSLNHPP to be full of people on 3-4 different teams desperately cramming in last minute miles so we don’t know which team is going to win until the last ride is uploaded.

Anyway, I’m still digging out from ignoring work while on vacation, but hopefully in the next few weeks I’ll have some time to dedicate to this.

[awitt]

I like the idea of modifying the rules so that more people can feel like they are contributing to their team’s success without having to ride hundreds of miles per week. I like the idea of a smooth function so people don’t have to get stressed when they miss their target tier because Strava recorded .2 miles less than Garmin/Wahoo reported. Calculus is helpful for this type of problem so I enlisted my husband to help because he does that sort of thing professionally. I started with a smooth function where points go from 1 to 0 as miles go from 0 to 100 which is y= -0.01x+1 the integral of this function, which gives you the area under the line for any value of x, is y = -0.005x^2 + x. There are many variations depending on whether you want to add a bonus at the first mile and whether you want the first mile to be bonus only or bonus + 1. The bonus can be used to change the relative weights of the early miles compared to the later miles. The attached image shows 3 variations. The % max column shows the percentage of total available miles for a ride that you get at each mile.

[ATTACH=CONFIG]17261[/ATTACH]

[DismalScientist]

I would merely point out that calculus is not required to find the area under a straight line. All one needs is mere geometry.

[awitt]

Is there a single function that will give you area for any point on x? I wanted something that is easy to program and fast to run. I was never good at geometry.

[Steve O]

@NemaVeze 176561 wrote:

Genuine question: what statistical test do you use for that? I’m trying to up my quant methods game.

Far from perfect. I run 10 sets of randomized teams and compare the highest and lowest scores. I don’t look at internal distribution.
This year ended with the 1st place team 74% higher than the last place team. I haven’t run a randomized sample against the final standings, but I probably will at some point. 74% is a pretty big spread, so I’m guessing that random will beat it sometimes, but less than half.

It’s this large spread (which is similar year to year) that have had some people propose adjusting the scoring system to make it tighter and hence more fun. This thread contains some ideas.

[chuxtr]

I think we should make this more like ice skating or diving. Throw out the high and low scores and award style points.

[Author]

Posts