Sudbury results are up:
http://www.dragonboat-truerank.blogspot.com/

I likely won't be answering any questions for the next little bit as I have lots of festivals to add.

Cheers!

Thanks again PD.

For the regattas that had distances other than 500m, do you find that the 100m/250m/1000m/2000m results are inline with the 500m results? I just wonder if there should be separate rankings for each distance as there would be in track for example where it's unlikely the 100m/200m/400m/800m events would have the same ordering of runners.

For the most part, crews 500m times are a very good predictor of all other races. As a result, the rankings would likely not be that different if say 200m results or 1000m results were used. However, even slight differences in time (say 1 second) can move you up and down the rankings quite far depending on where you sit.

Off the top of my head, only two teams have ever jumped out as being radically different (i.e. the regression lines everyone else close to a straight line and there is one dot way off the line). These two teams are not very surprising.

1. Eat Trout went much faster than expected for distance events last year.
2. Mofos go much faster than expected for sprint events this year.

I may come up with some kind of sprint/distance O'meter at the end of this year to indicate whether a team is a better sprint or distance crew.

Power Demon

I am sure this was explained previously, but I don't see it. How are the arrows derived? Are they in reference to a teams move up or down the rankings or is it related to their increase or decrease in speed year over year?

I see teams with slight increases in avereage speed with a down arrow and many teams with over a second faster average all with <-> "no change" arrow.

Still can't find your logic that keeps Mayfair in at 2.
Love to show the forum how you reason that out ?
Pickering & Welland is where you base the results ?
You must be on their crew.

Congratulations.

You were the only one smart enough to figure out the diabolical plot to put Mayfair in second place. We may never figure out how PD did it - what, with his "publicly available results" and his "reproducible methodology". He almost got away with it! But you figured it out.

Moron.

The arrows reflect the column "adj. chang". This is the year over year time, adjusted by the average improvement of the top 100 teams in the GTA ranking. In simple terms, think of it as your year over year change in time as compared to other teams.

...sigh. You had to know it was only a matter of time before people started accussing the rankings of being rigged. So yes, as laid out by Jerk Store, I take the time to doctor the regressions by adding fake coefficients and fudging times just so I can move my team up a few spots. If you are really worried about them feel free to replicate them, I will even send you the data and explain exactly what I did. I have done my best to make the rankings credible and unbiased, I simply follow a set of rules, pump in data and post the outcome. I really can't do anything more to improve their credibility.

Anyways, to answer the question. Mayfair is second since:
1. Based on Pickering their average time was between G&G and Senior National Team. When adjusted using the regression we get (based on three races).
G&G 1:58.9
Mayfair 2:02.0
Senior Nat 2:02.1

2. Using the adjusted Welland time (for four races) - don't be fooled by the slower times in the slower boats.
Mayfair 2:01.7

Final time = 02:01.8

On a more positive note, I will be providing crew times by festival a little later on.

Power D

Might want to consider adding lane data for every festival and using that data across years to help tweak the regressions and remove the lane bias argument.

Again, explain how you end up with a 2:01 time from Welland in the slower boats ?
Just curious to the analysis - so a 2:16 translates to 2:01 in heavier boats ??

I imagine lane bias is harder to quantify... and then the degree of the bias could be affected by so many other factors.

Someone needs to go out and make a huge indoor course - where did they film the movie Waterworld again??

It would be fantastic to remove lane bias but as Rob suggested likely quite the undertaking. I have kept the lanes in the data and may do a little look at how you might remove lane bias a later on. However, for this year at least the ranking metthodology will remain the same.

With regard to the Mayfair Welland time, first you should read the methodology if you have not already done so. It is kept at the bottom of the rankings page.

In brief, imagine that we are comparing two festivals. Of these two festival 15 of the same dragonboat crews compete in both. Now suppose that in the second festival all crew times are 3.0% slower. You can make two assumptions here:
1. Teams have all gotten 3.0% worse since the first festival.
2. Teams aren't slowing down but in fact conditions (boats, headwinds, course length) are actually worse.

The true rankings assume the second, and calculate the average change in conditions for each festival. It gets a bit more detailed as I use regression analysis to calculate the average conditions but just read the methodology and post if you have further questions.

Welland Adjustment Regression
(-20.0245274315299*(Time)+ 0.920245825103959)*Time


Mayfair Raw Times

Mayfair Predators Mixed 02:16.9 1 1000m
Mayfair Predators Mixed 02:16.3 1 500m
Mayfair Predators Mixed 02:17.3 1 500m
Mayfair Predators Mixed 02:17.5 1 200m

After plugging times into regression above

Mayfair Predators Mixed 02:01.7 1 1000m
Mayfair Predators Mixed 02:01.2 1 500m
Mayfair Predators Mixed 02:02.0 1 500m
Mayfair Predators Mixed 02:02.1 1 200m

Cheers,

Power D

I had no idea that the times were adjusted by almost 15 seconds for Mayfair at Welland(because of boats, yikes) yet there is nothing done for the Island (water level).Not sure if I agree with it but understand why you had to do it. However, it does put their ranking in the questionable zone. Not taking anything away from them being a great team.

Pickering used BuKs. So did Welland. So there are two fleets of BuKs, one of which is 15 seconds slower than the other? Or did they use some other kind of boat in Welland, maybe those old teak ones?

July 28 2006, 8:22 AM - No offense but I think you need to read up a bit on regression analysis. Absolute times don't matter when you use regressions. As PD has mentioned before he uses a base festival and regresses all results to that festival (I believe it is Pickering)

Previous poster - Welland used the old boats with the beam running down the middle. Much heavier and I don't believe they are made by BuK.

After all the recent discussion I went and read the methodology.. Nice. Very clean and simple, doesn't try to do too much and create significance out of nothing. I almost think that due to the small sample sizes it's better not to include lane bias, in general. Maybe for very specific instances like lane 1 vs. lane 6 at waterloo or lane 6 vs. lane 1 at Montreal... so then you would adjust to the middle lanes right? At a regatta if teams have about 4 races and the distribution of the lanes for each team is generally different from team to team........... well anyways the sample sizes become a lot smaller. But who knows if you can get it to work out then all the better.

Anyways my point is: good job!

Will there be a ranking for the Montreal festival?

What about Waterloo?

What about St. Catherines?

The actual base festival is the TIDBF (It was the biggest early season festival last year and for consistancy I didn't change it). As a result, TIDBF times do not get adjusted, but all other times get adjusted to it. Therefore, the rankings try an reflect how fast crews would go if they raise in the TIDBF boats, with the average TIDBF conditions, at the time of the TIDBF.

I think that the savey Dragon boat analyst does make some subjective tweakings to the rankings. However, the rankings are meant to reflect what, on average, has occured on the water. The rankings measure performance, not what we might guess crews would do if they all brought their best boat to a head to head competition.

As for a "questionable ranking" for Mayfair, I gotta say this seems really left field. The only competition Mayfair has seen was in Pickering. They came second overall (to G&G), and had the second fastest time on the day (to G&G). Other than Chiro (a good two seconds off in the ranking), all the other top ranked teams within 3 seconds were there. Clearly a second place ranking. Assuming the team didn't fall to pieces in Welland (which the data suggests they did not - nor would we have reason to believe it unless you have inside info), then why would they fall from second place? What am I missing?

We all know there are some crews out there that can and will go after them at the worlds, but the rankings are meant to reflect what has happened.

The lane bias idea is an interesting one. The sample size might actually be quite large in a sense, since all teams would have races in a many different lanes. I just don't have a clear idea how to do it at the molment.

Festivals to be added:
Montreal - This weekend.
Waterloo - Maybe, definetly have enough teams, a little worried about the lane bias, want to look at it, lower on the priority list than Montreal and Ottawa
Ottawa - Hopefully soon, waiting for data
St. Catherines - Not enough identifiable crews, sorry

Power D

Power Demon, you deserve some sort of award for all the work you have done!

This is absolutely amazing. The time you spend replying to everyone's concern and justifying your methodology is truely something to look up too.

Keep up the good work!