EuroClash, Tomodachi and Quad Wars, Three Tournaments, One Weekend (stats)

Outside of the USA, there were arguably three big “single date” tournament events last weekend:
Newcastle Roller Girls’ EuroClash, hosting European leagues, a WFTDA Recognised Tournament
Okinawa and Japan’s Tomodachi Derby Tournament, hosting East-Asia/Pacific and Alaskan leagues, with a WFTDA Sanctioned Tier
Mendoza, Argentina’s Lado Oscuro Roller Derby’s Quad Wars, hosting B/C level Latin American leagues.

There’s a lot of games in all of them, so there’s plenty of meat for statistical inference.

Let’s start with EuroClash.


EuroClash was arranged as a pure invitational tournament, with structure set mostly by the participants’ desires for opponents. As a result, EuroClash did not post a “winner” of the tournament. All bouts were WFTDA Sanctioned, as befits a WFTDA Recognised Tournament.

The participating teams for all but the last fixture were: Newcastle Roller Girls, Leeds Roller Derby, Middlesbrough Roller Derby, Tiger Bay Brawlers, Central City Rollergirls, Bear City Roller Derby (Berlin), Paris Rollergirls, Auld Reekie Roller Girls (Edinburgh) and Dublin Roller Derby. (Interviews with the teams can be found on this very blog here: )

The scores (from FlatTrack Stats) are:

TBB [Tiger Bay Brawlers] 161 CCR [Central City] 167
BCRD [Bear City] 228 NRG [Newcastle] 94
ARRG [Auld Reekie] 194 MRD [Middlesbrough] 99
LeedsRD [Leeds] 94 DRD [Dublin] 312
NRG 76 PRG 238
BCRD 177 ARRG 176
PRG 192 DRD 85
ARRG 301 LeedsRD 93
TBB 117 PRG 203
NRG 135 CCR 177
MRD 238 BCRD 89

As commentators have noted (including The Derby Apex), there’s some significant departures from prediction here, which we’ve highlighted in two different ways. Bolded results are surprising in both European and WFTDA rankings terms, whilst Italic are surprising in terms of WFTDA rank only [in this sense, we mean explicitly “from the WFTDA rankings, not from FTS’s ranking based on WFTDA Sanctioned play].

It’s pretty clear that there are several patterns here: two ranking errors in WFTDA (Leeds Roller Derby inheriting the ranking from Leeds Roller Dolls, and Paris Rollergirls’ known underranking) are responsible for the majority of the `surprises’ between them. Anyone who had seen LeedsRD play in British Champs would have expected the same high individual-skill, but teamwork needing more time to gel playstyle we saw in EuroClash – which is to say that there were moments of brilliance, particularly from their jammer rotation (Killaroo, Shin Pain and Little Dark One standing out for us, and in the stats), but there’s still some place to go with the cohesion.
And Paris Rollergirls’ determination to reach Division 2 playoffs this year is backed by considerable skill and effort in training over the last year – and a consciously constructed schedule of fixtures designed to gain them WFTDA ranking as quickly as possible.

In this context, then, the real surprises are: Tiger Bay Brawlers holding Paris Rollergirls to their expected performance [including holding them to zero points for the first half of the first period], Central City, in turn, ‘s surprising win over TBB in the first bout of the tournament, Auld Reekie Roller Girls matching their European expectations of tying Bear City to a 1-point win…

…and the curious case of Middlesbrough Roller Derby. The MRD bouts are the only bouts which are a really surprising result in terms of FTS – in the one case, MRD massively underperforming versus Auld Reekie, and in the second MRD massively overperforming against Bear City! (And we mean massively in both cases – Middlesbrough might have been expected to beat Auld Reekie with that kind of points margin, not lose to them by it!)

The possibility, exists, of course, that it was the other teams which were unpredictable… but we can test that with the power of statistics.

We performed a standard linear regression against the results from EuroClash, using score ratios as our measure of performance. As we’ve explained before, this assumes that the strengths of each team are simply related to each other by their performance – so if Team A is twice as strong as Team B, and three times as strong as Team C, then Team B versus Team C should give a 2/3 ratio win for the B team. Of course, performance varies from game to game, but we can use an “R-squared” value to measure just how good the approximation is for our results. A value close to 1 indicates that the results are very well described by this kind of model, whilst a value close to 0 indicates that no such relationship exists at all.

Running the regression against the full set of results above gives an R-squared1 of 0.39, a disappointingly low value for Roller Derby tournaments in general (which tend to be around 0.7 or higher).

However, given that we know that the MRD results are inconsistent, we can rerun the regression on the results, minus one or both of the MRD games. In this case, we get an pleasingly high R-squared of around 0.93 – actually very high for a tournament.
This suggests that, whilst the results might have been surprising when compared with WFTDA rankings, the performance of the teams (other than Middlesbrough) across EuroClash was remarkably consistent.

For interest, then, the relative power-rankings (which should be in the ratio of the points scored between two teams) are:

Middlesbrough Roller Derby 184 *
Paris Rollergirls 100
Middlesbrough Roller Derby 76 *
Berlin City Roller Derby 69
Auld Reekie Roller Girls 62
Tiger Bay Brawlers 51
Dublin Roller Derby 50
Central City Rollergirls 47
Newcastle Rollergirls 32
Middlesbrough Roller Derby 31 *
Leeds Roller Derby 17

where the three entries for Middlesbrough represent their performance against Berlin (184), Auld Reekie (31) and the average of the two (76). That average is around where we would expect MRD to actually have performed if they were consistent, which is interesting.

Generally matching the results against the existing rankings produces the best matches if we assume that Paris, Dublin and perhaps Auld Reekie, are underranked, and LeedsRD are overranked, with smaller shifts for the others.

This also means that, if we were awarding a Championship to any team, it would be the entirely expected Paris Rollergirls!

Paris will be competing against Bear City only this coming weekend, so this will be a good test of their true strength…

Of course, the final event at EuroClash was the Team Scotland Roller Derby versus Team Ireland Roller Derby bout.
A rematch of the game played at Road to TBC, where both teams were in their training roster (and Ireland had barely selected that), where Ireland won – this time Team Scotland’s competition 20 were playing a more-experienced, but still training, Ireland. (Both teams had many skaters from Auld Reekie and Dublin’s rosters, respectively, as well as other teams.)

It’s fair to say that there was a significant difference in the result, with Road to TBC’s tough-but-convincing Irish victory replaced by a tight even score in at the end of the first period, becoming a strong Scottish victory by the end of the second.

The final result, 209 Scotland : 128 Ireland is stronger than the expected result if Dublin had played ARRG, emphasising the importance of the other leagues making up each National Team; however, at half time, the result was a near tie.

1For the statistically inclined, this is actually the adjusted R-squared, not the bare R-squared. We’re also using the F-statistic for these regressions, but it closely follows adj. R-squared here.

Click here: for Tomodachi and here: for Quad Wars

One thought on “EuroClash, Tomodachi and Quad Wars, Three Tournaments, One Weekend (stats)

  1. Pingback: EuroClash 2018: Newcastle Roller Girls’ tournament expands beyond the “Euro”. – scottish roller derby

Leave a Reply

Fill in your details below or click an icon to log in: Logo

You are commenting using your account. Log Out /  Change )

Twitter picture

You are commenting using your Twitter account. Log Out /  Change )

Facebook photo

You are commenting using your Facebook account. Log Out /  Change )

Connecting to %s

This site uses Akismet to reduce spam. Learn how your comment data is processed.