menu menu

Tessa Crean

Tessa's events


Tessa's rowIT ranking

Rather than being a fixed number rowIT rankings take the form of probabilities. They look at the chances of an athlete's ranking being a particular value. These are displayed as graphs showing rankings for a class and set of boat types.

The graphs
mean 1444
dev 155
seed 1134
mean 1350
dev 109
seed 1132mean 1363
dev 96
seed 1171mean 1395
dev 112
seed 1171

{tl;dr} an implementation of a Glicko 2 rating system

When you are looking at the graphs:

  1. The scales are:
    • Horizonal: ranking. This goes from 0 (always loses) to 3000 (always wins).
    • Vertical: probability. This scale converts to a percentage e.g. 0.5 = 50%.
  2. The ranking probability curve is the dashed orange line. For example, the odds it is super-high or super-low are very low. Whereas the odds increase for a value somewhere between.
    • The higher and narrower the graph, the more certain the system is of the ranking.
    • The lower and wider it is, the less certain the ranking is.
  3. The mean is in blue. This is an athlete's most likely ranking.
  4. The shaded bars show the distribution for 1 and 2 standard deviations. The deviation value is in the side key.
  5. The red line marks two standard deviations below the mean. This is the figure used for seeding.

Every athlete's rankings evolve over their career. Statistical analysis of past performances establishes the rankings.

Analysis happens within each boat type and class (age/ability/weight). Rankings actually change race-by-race. But they only get updated online occasionally e.g. when needed for seeding crews.

An athlete's most recent results have the greatest effect on the update process.

During an update the system looks at all race results within the update period. It compares them with the expected results (based on existing probability curves*). The system looks at who was in the crews**. And it evaluates how crews performed against their opposition.

Then the system adjusts the existing probability curves to better fit the actual results. This can involve adjusting the blue line of the mean. Or it can change the shape of the curve e.g. making it looser or tighter.

* for boats larger than a single the crew members' rankings can be statistically combined to give a probability curve for the whole crew.
** this is an excellent reason why it is important to make sure crew changes are properly recorded!


The primary purpose of rankings is to use for seeding crews. Though, inevitably, they also get used for comparing athletes.

Seeding regattas

Seeding is mainly used for heats at NZSSRA Championships; because those regattas have elimination heats. There, a random draw can create a situation where top crews get eliminated early.

For example: if only 2 crews can progress from a heat then a random draw can put the top 3 crews the same heat. That would knock out a medalist during the heats.

The goal of the seeding system is to prevent such undesirable draws.

This happens by getting as close as possible to an optimum draw. One where all crews are perfectly seeded. Yet it is not necessary to get seeding 100% correct. (If that was possible you could simply skip racing and hand out medals)

At last review Maadi seedings were about 85% accurate. The system was able to accurate predict the placing of 85% of all crews in every heat race. More than adequate for the job!

Athlete comparison

It is possible to use rankings to compare athletes.

But anybody doing that must bear in mind the fact rankings are super-specific. They relate only to the age-group and boat type shown. This means:

It is only valid to compare rankings for the same age-group and boat type.

Comparing rankings across classes is meaningless. And even relevant comparisons are challenging.

  • The easiest comparison is seeing which athlete has the higher mean.
  • But a more accurate comparison may be the seeding value. That takes some uncertainty into account.

In reality, the best way of comparing rankings is to race the athletes!

At the moment only the rankings for an athlete's actual age-group are showing. But rankings for other classes are also held (and are usable for seeding).

Log in

This version of rowIT is no longer maintained. Logins are only usable with the latest version of rowIT. Head to

Logging in will only make the full profile accessible if Tessa Crean is a member of your organisation.

June 18
KeyEntries open
  Entries close
Reserve day
Results published
map of New Zealand North Island Northland Auckland Waikato Bay of Plenty Gisborne Hawkes Bay Taranaki Wanganui/Manawatu Wellington South Island Marlborough Nelson West Coast Canterbury Otago Southland