|
Post by chileduck on Jun 1, 2019 12:32:25 GMT -8
I'd like to get some user reaction to a method I developed for showing form chart projections shifts as a meet progresses. Below is a series of screenshots showing the displays as the 2019 Indoor championship for the women progresses. There are links to a version of the formchart tracker so you can view it as if it were 'live' as well. I'd like to hear if this display makes sense and if not, what are your questions and/or confusions. Recommendations for changes are welcome but not necessary, as I'm mostly interested if you can make sense of it without further explanation than is what is displayed on the screen. So here is the progression: Before competition: Link to interactive versionAfter prelims: Link to interactive versionAfter Day 1: Link to interactive versionDuring competition Day 2: Link to interactive versionEnd of competition: Link to interactive versionLet me know what you think.... In case you didn't know, when you view the results in one of the mid-competition versions there is a "Compare to Projection" link at the bottom of the event list for showing displays like this:
|
|
|
Post by olduck66 on Jun 1, 2019 17:17:38 GMT -8
Chile Sort of taking the surprise out of not knowing which team is doing well and who has stumbled. I can guess the level of excitement when the Ducks are ahead of their predicted pace. We Duck fans for for sure have the top internet site in the nation. Thank you!
|
|
|
Post by gnorthern on Jun 1, 2019 19:19:13 GMT -8
We Duck fans for for sure have the top internet site in the nation. Thank you! Are there any other universities with a track forum like this? And thank you Chile.
|
|
|
Post by duckcardinal on Jun 3, 2019 12:09:15 GMT -8
Projection value seems to float from pre-meet to post day 1 prelims before it reaches its set point for final +/- valuation. This seems that it might give a skewed version of the accuracy of the original forecast scores. e.g. AR does great in prelims and lifts its projection from 71 to 74, but later under performs to only score 62, which is shown in the end of meet final as being -12, versus the -9 it would be from the original Flotrack forecast. Similar with Oregon & Ohio State, which is as far as I went. Am I missing something, or is that just a wrong cell reference in the final cell formula (or upstream)?
Regardless, this looks very cool and I'm looking forward to giving it a try!
|
|
|
Post by chileduck on Jun 4, 2019 7:49:55 GMT -8
Projection value seems to float from pre-meet to post day 1 prelims before it reaches its set point for final +/- valuation. This seems that it might give a skewed version of the accuracy of the original forecast scores. e.g. AR does great in prelims and lifts its projection from 71 to 74, but later under performs to only score 62, which is shown in the end of meet final as being -12, versus the -9 it would be from the original Flotrack forecast. Similar with Oregon & Ohio State, which is as far as I went. Am I missing something, or is that just a wrong cell reference in the final cell formula (or upstream)? Regardless, this looks very cool and I'm looking forward to giving it a try!
Thank you, duckcardinal, for taking the time to look at this carefully. And you answer the question I had about whether the method I chose would make sense. What you are seeing is a result of adjusting the projection after prelims and using that adjustment to calculate the final +/-. So using your AR example I calculate the final -12 by comparing the final score with the post-prelim 74 projection rather than the original 71 projection because I thought it would tell the story better because the expectations change for Arkansas (and others) after preliminaries and wanted to reflect that yes Arkansas DID lose 12 expected points off the form chart. Really I'd like to show +/- compared to both and I guess I can come up with a way of making it clear which is which but that would take some more design work and not sure the distinction is easy to make. Perhaps it's better I just choose to use the original projection for calculating +/- because that is what you and probably most of the more casual fans than you would expect. I made an attempt at this showing both with the comparisons I show for individual events. I don't know if anyone realizes it but if you look at the results of events that have preliminary rounds and click on the 'Compare to Projection; link and the bottom, it gives you a tabbed panel so you can see both... Below are some screenshots showing this. For now I think I will change this to just use the original projection to base the +/- calculations. Look back later and I'll recreate the flow example using this method and you can tell me if it makes better sense. Also let me know if you think it would make sense to use both if I can figure a way to let the user choose and make it clear what they are choosing. Screenshots: [ In looking these over, I think it will probably be best just to use the original projection in the +/- calculations.]
|
|
|
Post by chileduck on Jun 4, 2019 9:01:54 GMT -8
|
|
|
Post by bruce3404 on Jun 4, 2019 10:19:08 GMT -8
Would be interesting to run the various projected numbers from FloTrack, TFN and even USTFCCA over a lengthy period (which would be limited by how long each of the aforementioned three have made such NCAA projections) and see which format has been consistently most accurate. One caveat would be that for FloTrack and TFN, they've undoubtedly used different humans to come up with the numbers; I suppose it might be fun to run a battle of TFN's Auka and Pfeiffer (both of whom have been doing this for quite some time) vs FloTrack's analysts and then use the winner for your charts. I'm sure you've got 100 hours to spare doing this
|
|
|
Post by duckcardinal on Jun 4, 2019 15:09:25 GMT -8
Thanks Chile (and Bruce); after considering your comments I've come to think that it depends on what you want to see, and when you want to see it.
If you're keeping psychic score on which formcharts prove the most robust from pre-meet to final results then the tweak you've made would be best for that. However if one is looking to play along in detail, then the rolling adjustments are actually more informatively rich in catching swings of momentum, providing more granularity in various outcomes, so it would probably be more fun to "play" given the considerations Bruce presented Re effort in prelims, prior performances etc. Hope that makes sense.
This should provide some interesting forensics after the meet. Cheers guys!
|
|