Most "predictive" event?

[SPP SciO]

Just curious: are there any events which tend to be better predictors of team success than others? In other words, events where the medalists are more likely to be part of a top-placing team than medalists in other events.

My first guess would be the reading-heavy events (disease detectives, for instance) because those skills translate most universally. And, build events would be poorer predictors, since the skills are so specialized. But I'm pretty sure it would be easy to run some sort of analysis on this (has it been done?) Real science is welcome here but feel free to share if you just have a "hunch"

[windu34]

Just curious: are there any events which tend to be better predictors of team success than others? In other words, events where the medalists are more likely to be part of a top-placing team than medalists in other events.

My first guess would be the reading-heavy events (disease detectives, for instance) because those skills translate most universally. And, build events would be poorer predictors, since the skills are so specialized. But I'm pretty sure it would be easy to run some sort of analysis on this (has it been done?) Real science is welcome here but feel free to share if you just have a "hunch"

There are definetly patterns. Schools such as Poudre and Harriton are strong build schools and are known for doing well in builds and nats. On the other hand, scools on the east coast such as Columbia and Fayetteville Manlius are known for being strong in Dynamic planet, Green generation, etc.
Then you have Troy, Solon, LASA which do well at everything (except for chance events such as widi). Essentially what we would call the "outliers"

[AlphaTauri]

I did this analysis a while ago for a few years of PA States (Div C) -- the link is here if anyone wants to check it out.

The statistics I used were pretty basic, a chi-square test for association/independence and the r^2 value of overall place vs. event place (because I am not a statistician and those seemed like the best options of the ones I knew).

With the important caveats that of course events and event supervisors may change from year to year, and that these are of course only general trends, here are my results:

• Of the perennial events, Astro and DynPlan seem to be the best predictors
• In general, good predictors are usually ID events or "hard"/physics-y events
• Some events (ExpDes is the most egregious example in my dataset) swing back and forth between being good and awful predictors, my guess is that this is due to different supervisors with different styles
• Worst predictor appears to be WIDI for... obvious reasons
• Builds are generally not good predictors either, which is also not very surprising

If anyone is interested in this stuff and/or knows more about statistics, feel free to play around with the spreadsheet.

Edit:
I have a hypothesis that if you tried this analysis for both Div B and Div C, Div B events would correlate less overall because there is generally more parity between teams.

I'm also not sure if Nats would be the best dataset to analyze, as there is a VERY wide gap between the best and worst teams, which means the event placements generally tend to split into tiers with each team more or less staying in their tier.

[syo_astro]

I feel like someone else has analyzed this on the forums before for some competition, probably nats. My bias of course would be to say astro . I do believe usually the best schools don't always dominate across the board at builds and inquiry events compared to studies just because the competition, environment, travel, who knows what can provide inconsistency. Sure some schools do a bit better at builds, but considering there's so few it's hard to find teams that medal in EVERY one year to year. Though, then there's those times where you get a person or two that wants to be an engineer, really into scioly, etc...wrecks the competition!

[varunscs11]

Just curious: are there any events which tend to be better predictors of team success than others? In other words, events where the medalists are more likely to be part of a top-placing team than medalists in other events.

My first guess would be the reading-heavy events (disease detectives, for instance) because those skills translate most universally. And, build events would be poorer predictors, since the skills are so specialized. But I'm pretty sure it would be easy to run some sort of analysis on this (has it been done?) Real science is welcome here but feel free to share if you just have a "hunch"

There are definetly patterns. Schools such as Poudre and Harriton are strong build schools and are known for doing well in builds and nats. On the other hand, scools on the east coast such as Columbia and Fayetteville Manlius are known for being strong in Dynamic planet, Green generation, etc.
Then you have Troy, Solon, LASA which do well at everything (except for chance events such as widi). Essentially what we would call the "outliers"

To be fair, LASA did get 13th in WIDI in 2015 which I have to say amazing job to the people from my school because it was the first time they actually did WIDI together since we didn't compete in WIDI and State and Regionals. Also Texas swept Green Generation (1st-Seven Lakes, 2nd-LASA). Although I do agree with you. Hawaii does amazing on geology events and Texas does pretty well on them too considering Texas's unique geologic history. Just wanted to throw this in too but Solon has an AMAZING engineering team. You could argue that they do better on engineering as opposed to study while Troy does better on study events. But the reason Troy and Solon have won so many national tournaments is because they are extremely consistent across the board.


I would say the most predictive event is Disease because usually the people who win disease have been doing it for a while and consistently win because Disease is so speed based. For example Camas and Centerville won disease back to back. Events that are pretty consistent are the events where once you learn the material you are good aka the events where you don't have to continually study (i.e. not anatomy not cell bio not entomology not TPS etc)

[Unome]

I feel like someone else has analyzed this on the forums before for some competition, probably nats. My bias of course would be to say astro . I do believe usually the best schools don't always dominate across the board at builds and inquiry events compared to studies just because the competition, environment, travel, who knows what can provide inconsistency. Sure some schools do a bit better at builds, but considering there's so few it's hard to find teams that medal in EVERY one year to year. Though, then there's those times where you get a person or two that wants to be an engineer, really into scioly, etc...wrecks the competition!

I think someone did an analysis for last year's Nats, although I can't remember exactly.

[bernard]

We did Spearman's Rank Correlation Test on 2014 Nationals results.

[watermydoing14]

Just calculated correlation between team rank and event rank for each event at the past 2 Washington State competitions.

The top four "predictive" events were, in order: It's About Time, Wright Stuff, Astronomy, and Green Generation.
The least four "predictive" events were, from least predictive to most predictive: Scrambler, Water Quality, WIDI, and Experimental Design.

The only build events that had a correlation of more than about 0.6 were Wright Stuff and Elastic Launched Glider, so there's something about flying events that make them more predictive. Seeing as other predictive events were It's About Time and Astronomy, both events that you can bring an entire binder or laptop, my theory is that teams that place higher generally put in more work, therefore, their binders/laptops will be more extensive/better, then they are more likely to do better in those events even if their studying and ability to memorize is not super good. As for the flying events, that likely has to do with the large amount of testing that is needed in order to do well in the event. Teams that rank higher are probably more likely to put more time into testing their planes/gliders, and consequently are more likely to place higher in those events.

All the events fell into a correlation range of about 0.3 to 0.75, except for Scrambler, which had a correlation of 0.019 ( ). I have no experience in this event, so I have no idea why that would be, but I know that in Washington state 2014, that entire event was DQ'd, but I didn't account that into my calculations so maybe the event itself had problems in that things were too likely to go wrong, so teams had a relatively high chance of something getting screwed up at the last minute, and that chance of something going wrong would be even for all the teams, so there's a good chance for a team that ranked lower overall to do much better than a team that ranked higher overall if the team that ranked higher just happened to have something go wrong with their device.

To convert this into strategy: You are more likely to have to put more work into placing higher in more predictive events, so it's more important to put people that you know will work hard or people with the most experience on those events.

[meteorology891]

One would think ID events like Ento or Invasive are predictable because they have lists, but you have no clue what they will ask you about that species. I went to competition for Invasive and got the same 10 questions for 11 different species.