Decathlon statistics

[olorin]

Finally a comparison between the performances in the 100 compared to the two technical events PV and JT. Like the previous table the relative performance is divided to five and then I calculate the relative performance in these two events.

	PV	JT
1	-8	-29
2	-14	-21
3	-15	-23
4	-12	-11
5	-9	-3

The only explanation I can come with is that decathletes treat the PV as a different animal and their success is more about technique then their overall abilities. However, I would except then the JT to be much the same. Any other ideas are welcome.

[olorin]

I assume you're already scraped the results from the top lists at the worldathletics.org page? They appear to be missing many performances, however. They have 1638 performances over 8000, but there have been about 2500.

And the results they miss are typically old one.
There are three sites that I use:
The most comprehensive one is decathlon2000 - but it is also the least friendly to extract data from.
All-time athletics is (as always) wonderful. Of all 8,200+ scores there are four or five results they miss. The problem is that the site only has the overall results and not the individual events. When I will finish my updating (probably in 2023) I will offer my data to Peter Larsson.
World athletics has so many omissions (and mistakes) that it is really pathetic. To illustrate, from 8140 to 8100 there are 288 performances, of which only 164 (less than 60%) are in world athletics.

[olorin]

Results:
As previous analyses, I divided the relative 100 into equal groups only this time I used 5 (quintiles) as the more interesting results (IMO) are in the middle three quintiles rather than the two extreme one. Then I calculated the average relative performance in events 2-9 (LJ through JT). In the next two columns I split this performance the three sprint events (LJ, 400, 110h) and the other six.

Group	Relative 100	Relative 2-9	Rel_sprint	Rel other
1	Below -56	-33	-55	-21
2	-56 to -32	-21	-30	-16
3	-30 to -11	-16	-21	-13
4	-10 to 10	-10	-12	-10
5	Above 10	-4	-4	-5
4-2		11	17	6

The first columns is the group number (1 is the worse). The second column is the relative performance in the 100 (see technical notes). The next three columns are the important one. Column three is the average relative results in the next nine events (LJ-JT). The final two columns are splitting the nine events into sprint (LJ, 400, 110h) and the other five events (SP, HJ, DT, PV,JT).

When I started the test, I expected that the two extreme quintiles will have the effect that I found. I have to say I am more surprised that there is relatively large gap between the middle quintiles (see last row). Note, that the numbers are averages so a difference of 11 points is actually 88 points in the next eight events.

[olorin]

Technical notes

As previous questions, I concentrated only at the performance of decathletes with PB of 8200. That is, I include only performances that the decathlete had a PB of 8200+ at the start of the decathlon. Due to data limitations (from decathlon2000 site) only decathlon with a final score of 8,000 points are included. Additionally, many of the old timers are not included. The total number of performances that I analyze is ~1,000.
Of these I compared between two performances. The performance in the current decathlon vs. the performance in the decathlon that the athlete set his decathlon PB (at the time of the performance). I first calculated the difference in the points in the 100. The average difference in -24 points, which suggests (not surprisingly) that decathletes typically run 0.1 slower than the time they ran when they set their decathlon PB. Then, I calculated the difference for the next eight events (ignoring the 1,500) and examine the relation between the two.

To illustrate what I have done, let's look at an hypothetical decathlete Kevin that have five performances above 8,000. Each of these performances is divided to three parts. Score in the 100, score in the next eight events and the score in the 1,500. The following table presents these performances in chronical order.

	Total score	Score 100	Score 2-9	Score 1500
1	8100	950	6,400	750
2	8250	1000	6,470	780
3	8050	1050	6,250	750
4	8400	1030	6550	820
5	8150	1045	6400	705

The first two performances of Kevin are ignored as at the time when he starts the decathlon he didn't have a PB of 8,200. After the second decathlon when Kevin score above 8,200 he enters the analysis and his performance in decathlon 2 becomes the benchmarks. In decathlon 3 Kevin score 1,050 points in the 100 compared with 1,000 in his benchmark, hence his relative mark is 50. The result in the next eight events 220 points lower. I divided this number in 8 to calculate the average relative performance in these eight events (from LJ to JT) so the score is -27.5. The following table presents the calculation for Kevin's decathlon performances

	relative 100	relative ave. 2-9
1	N/A	N/A
2	N/A	N/A
3	50	-27.5
4	30	10
5	15	-18.75

Note, that after performance 4 when Kevin broke his decathlon PB, this becomes the new benchmark that performance 5 is compared to.


Lastly, I ignore the performance in 1,500. The reason behind it is that the performances in the 1,500 is often related to the relative performance of the decathlete rather than the shape that he is in. In other words decathletes will be willing to punish themselves more when chasing a medal then when they are not in contention.

[olorin]

So, you had a good start – so what?

The last question before the Olympics (T&F) start is concentrating in the importance of a good (or bad) start to the overall decathlon score. In other words, can we predict the results of the other events based on the score in the 100.
In order to examine this question, I look at the performances of all decathletes that had a PB of 8,200 and above. For these decathletes I examine the result in the 100 and compared it to the result of the 100 when they set their PB in the time. Then, I examine their performance in the next 8 events compared to their performances when they set their decathlon PB. Because of data availability I have only performances above 8,000 points.

The bottom line
The performance in the 100m is correlated with the next 8 events.
The relation is much stronger in bad scores than in good scores. A really bad score in the 100 on average suggests a decrease of ~150 point in the next eight events compared with an average one, whereas a good performance suggests an increase of ~90 points.
The relation in stronger for the other three sprint events (LJ, 400, 110) than the other 5 events. While wind may have being a possible reason behind this result the relation between 100 and the 400 is relatively very strong.
No relation between the performance in the 100 and the performance in the PV.
Finally, the relation between the 100 and the other day 1 events is stronger than day 2 events.

[tgs3]

One of the question I plan to ask is whether decathletes changed through the years.
The problem is that the number of observation before 1984 is relatively small (roughly 100).
I'm trying to increase the number by collecting the data for lower marks (the ultimate goal is 8000). If anyone has the data and not mind to share it, then I promise a quick answer.

I assume you're already scraped the results from the top lists at the worldathletics.org page? They appear to be missing many performances, however. They have 1638 performances over 8000, but there have been about 2500.

[Trickstat]

HJ and JT is a combo I'm not particularly aware of. However, 1984 Olympic women's JT champ Tessa Sanderson was a 6000+ heptathlete and actually first competed at the English Schools' Championships in the high jump. However, her second best heptathlon event was probably the 100H where she came second in the English champs in 1981. In fact, I have heard people talk of a correlation between hurdles and javelin. However, I think that is more of the case that good hurdlers are quite often decent javelin throwers rather than the other way around. I know Colin Jackson was nationally ranked in his age-group as a javelin thrower in his early to mid teens.

[olorin]

I'm kinda with Dwight. In my experience/observation, back in the olden days, when many/most athletes did as many events as they had some aptitude for, HJers routinely doubled as JTers.

As I said I have a limited data on the "old days".
Even, when I limited the data to before 1990 the negative relation between HJ and JT continued.
The negative relation reduces a bit when I limit the sample to 1985, but with so little observations it is really hard to tell whether this is a genuine reduction or luck.

[Fortius19]

I'm kinda with Dwight. In my experience/observation, back in the olden days, when many/most athletes did as many events as they had some aptitude for, HJers routinely doubled as JTers.

A lot of Deca training advice recommends training the two events together on the same day due to similar run-up patterns.


This is great stuff Olorin! Thank you. I remember back in early internet days, some random website did a semi-scientific research of the Deca events. They determined the Discus was the #1 key to a good score and gave a descending order of importance. I took notes somewhere and misplaced them. I haven't been able to find the website since. I do believe the PV and JT are key technique wise.

[lonewolf]

If you watch the pattern of run and take off, the two events align well. Though the great HJers don’t generally throw well, the good Jav throwers HJ relatively well.

I'm kinda with Dwight. In my experience/observation, back in the olden days, when many/most athletes did as many events as they had some aptitude for, HJers routinely doubled as JTers.

[noone]

If you watch the pattern of run and take off, the two events align well. Though the great HJers don’t generally throw well, the good Jav throwers HJ relatively well.

I have two words for you: Rory Kotinek

[Atticus]

surely you can tell the system that those with more than 500 posts should be off the suspect list.

Don't feel so bad; I have a bazillion posts and I get 'suspected' all the time. It's usually due to a link in my post.

[deca]

Olorin, do your calculations include pre 1984 performances? Any incite on how modern decathletes compare with the guys from the 80s and 90s would be appreciated. With my untrained eye it looks like the guys filling the top spots on the all-time lists, your Ashton Eaton’s, Damian Warner’s and Kevin Mayer’s are all speed based athletes and therefore it’s becoming harder to be truly world class as an all-rounder or throwing specialist. Pappas, Harder and Clay were exceptional, but it was Eatons extra speed that took him to the world record

or was it Eaton was tough on the long runs. Clay’s PBs on speed events aren’t way off. He just never ran a great 15 and kind of never had to. If he’d run 414 (800) instead of 440 (650 Ish) he would have scared 9k at the OT in 2008. In Dec, it ALWAYS comes down to an event or two. Whether we are lousy guys hoping to get a medal at junior Olympics, or the guys winning gold in the real thing.

It is always just about an event or a bar or a day.

[deca]

Stones used to say that high jumpers tended to be good javelin throwers. (Maybe he still does, though I haven't seen him cover a deca in a long time.) Any evidence of that?

If you watch the pattern of run and take off, the two events align well. Though the great HJers don’t generally throw well, the good Jav throwers HJ relatively well.

[cigar95]

. . . . . . Turnout that the flukes happen in reality much more than I thought. I don't think that this mistake suggests that Stones is not knowledgeable, I have no idea (I an not American so I hardly heard his commentary). I think that he made a mistake that if often done by people (including, coaches, athletes, experts and statisticians) because we are human.

My guess is that Dwight remembered some examples of good HJ / JT combinations, and perhaps a couple bad pairs, and extrapolated based on a small data set in his memory.

Particularly interesting is how little the vault is correlated with much of anything, other than that Ryan Crouser is not going to jump 5.80. (Yeh, I know he's not an 8200 point guy, I'm just hyperbolizing . . . ). I would guess that this speaks to the importance of technique in this event, given that everyone is already very physically talented. (Plus, we see that the JT has only mild correlations to anything else, going back to the earlier discussion in these events.)

I was also going to remark on how strongly the matrix correlates with its transpose, but then that almost has to be the case. If good 400 runners are lousy discus throwers, then good discus throwers are going to be lousy 400 runners.