Thoughts on Fairness in Rowing

I’m sure many of you have closely followed the olympic rowing regatta. One of the highlights of the event was the close finish in the men’s single sculls between Mahé Drysdale of New Zealand and Damir Martin of Croatia. The photofinish turned out favorably for Mahé, winning him his second olympic gold medal in the men’s single sculls.

How large was the winning margin? Half a bow-ball. That is approximately two centimeters. That is easily the difference between rowing just a tiny bit more efficient or not. Dealing with rowing performance measurements all day, we know how important it is to consider the accuracy and precision of the measurement equipment when interpreting data. How does this apply to the olympic racing? Our exploration starts with a simple question: who should win this race?

drysdale_photo_finish-large_trans++tR1WRZ4lb6BOb1-FBRpn-lFGGjcI28awSNtOPAVs6pc

The rules of rowing are pretty clear. A race is won by the crew that crosses the finish line first. Nonetheless, this close result sparked a discussion why the race was not declared a tie and both participants were awarded a gold medal. After all, as spectators we frequently see ties in other disciplines like swimming, a point I will come back too at the end of this article.

The Photofinish

As much as the decision of the jury was debated, it was absolutely correct. Let’s have a look at the relevant regulations in the FISA 2015 Rulebook, APPENDIX 4 – Bye-Laws to Rules 42 to 44.

3.3.4.  Timing and Results Systems (p. 147)

Times shall be shown to 1/100th of a second on the Results Sheets and on the Scoreboard.

In the case of close finishes the order of finish must be determined by means of special equipment such as a photo-finish camera, capable of measuring and displaying differences to at least 1/100th of a second.

The official results reported both participants at a time of 6:41:34, which is a consequence of this rule. It’s not a limitation of the actual equipment used for the timing. The timing equipment used shows a clear difference in the photofinish and quite likely supports a much higher resolution of times than 100 frames per second (which equals 1/100th of a second). Would a “lesser” photofinish camera have awarded a tie? Let’s do some back of the envelope calculation. The boats moved approximately at 5m/s, or 5cm in 1/100th of a second. The winning margin was 2cm. So the answer is… probably. My take on the matter? It’s not fair to judge a race down to a precision the equipment can’t deliver.

I think it’s very unfortunate that the Rulebook does not state the exact precision of the photofinish equipment to be used. Whether a tie is awarded or not could (theoretically) depend on the photofinish resolution. For practical purposes though, this likely does not matter as long as the resolution of modern photofinish cameras is typically on the order of 1/1000 – 1/20000 of a second and very few races need to be decided on the level. However, and this is an important point, “infinite” precision would always be able to decide any race. But even then you need to stop at some reasonable level because the physics gets murky at the quantum level. To illustrate this point, let’s look at this photofinish of the BLW2x semifinal at the U23 World Championships in Rotterdam last weekend:

BLW2x-photofinish

Looks like a tie, doesn’t it? Here’s what the rule book says:

Rule 80 – Dead-Heats

When the order of finish between two or more crews is too close for any difference to be determined, then the result shall be declared a dead heat between the crews involved.

But imagine you could zoom into the image some more because a nanometer-resolution finish camera was used. It’s possible you could see a difference if you can see a couple of nanometers. Would it be arbitrary to say 1/1000 of a second is the resolution to be used (~0.5cm in most rowing disciplines)? Yes. But allowing use of anything better than 1/100 of a second  is very arbitrary as well.

What could FISA do? Specify the precision, make it slightly less arbitrary than it currently is.

Course Length

The Olympic rowing course is 2000m long (Rule 43, p-. 67). Quoting the rule book:

2.Stretches of Water (p. 139)

A standard international course shall be straight and shall not have less than 6 racing lanes. It shall provide fair and equal racing conditions for six crews.

The rule book even prescribes a minimum water-depth (which is very important as it has an effect on the drag of the shell in the water). However, the rule book does not mention how the length of the course is to be determined. Modern survey equipment can easily measure down to millimeters of accuracy, but can the course be built to those specifications? What is the allowable tolerance? And more importantly, what is the allowable tolerance between lanes? Unfortunately the rule book does not specify those.

In any case, from a fairness point of view the most important aspect of the course is that the start line is perfectly parallel to the finish line. The standard procedure is to visually align the boats from the aligners hut:

3.1.6 Aligner’s hut (p. 142)

This shall be a fixed structure placed exactly on the start line, ideally not less than 15 m from the first lane and no more than 30 m.

Again, the tolerance is not specified. How exact is, “exactly”? The aligner visually aligns the boats, which has its own problems if high precision is required. In an olympic regatta, the alignment is thus delegated to an Alignment Control Mechanism (ACM), sometimes also called “starting shoes”.

3.1.10. Alignment Control Mechanism (p. 143)
All regattas may use an alignment control mechanism in the center of each lane which shall hold the bow of the boat in a fixed position on the start line until the Starter makes the start. At Olympic and Paralympic regattas the use of such an alignment control mechanism of a type approved by FISA shall be mandatory.

How accurate is the alignment using an ACM? The system must be approved by FISA, which is good, but does “approval” cover a survey of the installation of the system as well? What precision of alignment can the system maintain under changing thermal conditions (e.g. a metal frame may expand), when loaded with wind? How rigid is the ACMs construction? These points may all seem like nit-picking. However, I feel they are important when we consider judging races by margins of two centimeters. I think it’s not fair to judge a race down to a precision higher than that of the course.

What could FISA do? Specify the allowable tolerances for course length and alignment control mechanisms. Make them part of the “accreditation” criteria for a rowing course. 

Well, but what about conditions then?

We accept “unfairness” in conditions between lanes all the times due to wind and waves, which may be different between lanes for various reasons like the famous “shelter island” around the 1000m mark in Brandenburg, this year’s host of the European Championships. Depending on wind conditions, “shelter island” gives an advantage to the lanes closer to the island or to the shore. If the effect of outside conditions is so important, why bother with course length and photo finishes?

I can very well understand this argument. The Rule book mentions the “nature of the sport” in a number of places. I think this fundamental “unfairness” caused by environmental conditions is part of the nature of rowing. Just like a possibly wrong decision made by a referee is part of the nature of soccer. But it’s easy to argue it would be unfair if one goal was slightly larger than the other one, even if it’s just a few millimeters. The same applies to rowing.  The vagaries of outside conditions are part of the sport, being imprecise is not.

How can we deal with this “unfairness”? If you can’t control it, make it part of the game. That’s what we do in rowing too. The lane assignment is based on the results of the heats and repechages. Each crew thus has equal chances to race for a “good” lane.

There was some debate about the “fairness” of the default lane assignment after the 2014 World Championships, when the German M8+ lost to Great Britain by a margin of 0.66s. By default, lanes are assigned “inside out” with the fastest crews from the heats receiving the center lanes and the slowest crews receiving the outer lanes. In Amsterdam at the time of the race, wind conditions were toughening up. The Germans had won their heat over GB, who had to go through the repechage. That’s why the Germans received lane 3 (center) while the British team received lane 5. Being closer to the shore, lane 5 was a bit better protected from the ensuing cross-winds.

Some commentators of the race argued, the jury should have changed the lane assignment in those conditions in favor of the German team. One rule that enables such an action by the jury and the starter is the Bye-Law 1.3 to Rule 96:

1.3 Unfair or unsafe conditions

The Starter shall consider whether the wind is likely to create unequal or unsafe conditions and, if directed as described below, or after consulting with the President of the Jury, shall take whatever steps may be necessary in accordance with these rules to ensure a fair and safe race. The President of the Jury shall inform the Starter of any required changes at least two minutes before a start.

From the point of view of the athletes involved in this race, the decision of the jury must seem arbitrary. What could FISA do? Why not make the lane assignment decided by the crews themselves or their coaches? This would immediately end all debates of blaming and criticizing the jury’s decisions around lane assignment. Let the starter have the crews draw their lanes by the order of their heat results, we already have a system in place for ranking those. I realize this may be hard to implement logistically, but I think it’s worth thinking about this.

Fairness in Swimming

Swimming (at least the indoor variant) does not have to deal with the vagaries of outside conditions. However, most distances raced are much shorter than in rowing. Unfortunately, most pools are built from concrete. Despite its apparent simplicity, casting concrete to exacting specifications is a very difficult undertaking. The international swimming association FINA thus states that the allowable tolerance for pool length is 3cm. That’s about the distance a swimmer travels in 1/1000th of a second (source). Consequently, ties are awarded if the finish falls within this margin of error.

Should rowing adopt a similar ruling? I think it would enhance the fairness of the sport. Let’s discuss in the comments!

24 comments

  • Johannes Lotz

    I think a more precise photo-finish camera would be the first step for the FISA! But the question about the precision of the ACM system is also a very good aspect! I can’t believe that the boats start exactly on the same level (with a precision of less then 2cm!!!).
    Your sentence >>I think it’s not fair to judge a race down to a precision higher than that of the course.<< sums the current issue up! Either the FISA modify their rules following the example of the FINA or they define their photo-finish camera and the ACM system!

    • Johannes

      Thanks for your comment and support! It’s not so much an issue that the photo-finish cameras that are actually used aren’t precise enough (case in point, they are likely accurate to about 1/10000 of a second or better). The issue I have with the current photo-finish rule is that the declaration of a dead-heat can depend on the resolution of the camera used (i.e. the race in rio could have been a dead-heat if the resolution was only 1/100 of a second).

      • St.

        Only the very new finish camera model by Omega (that’s the camera brand used for World Rowing events, [to be fair its actually swiss timing, but that’s another story]), i.e. the “Myria”, can operate up to 1/10,000th of a second. The one’s that are in the regular equipment since approx. the Beijing Olympics (OSV Star) “only” can take up to 2,000 images per second. However, the standard setup for rowing is to operate at 1,000 per second. This is likely to not change in the future as you end up with sharpness and lighting problems when you increase the camera speed and because of the low boat speeds in rowing (compared to short track, e.g.).

  • Kevan Armstrong

    All of what is said and presented is agreed my big concern is that all this trouble is not mirrored at the start where a human holds the boat into a shoe, are all crews level to the accuracy you can measure at the finish, it would never seem worth while checking if they were within a 1000th of a second level it would take to long to set races off, my point is don’t judge a finish by a more finite distance than you can garuntee at the start.

  • Ron Paterson

    Another major issue that affects the result in these circumstances is the point of the stroke cycle each sculler is on as his bow crosses the finish line. In this case, Mahe in fact appeared to be slightly behind Damir as the two scullers approached the line, but because he was at frontstops as he crossed the line his bow had surged ahead at just the right moment to give him the gold medal. This is not so much a question of fairness as one of luck, but it can make a much bigger difference to the outcome than matters such as the precision of the alignment on the start line.

    • Johannes

      Yes, the position in the stroke cycle is important. See also this visual introduction on how Rowing in Motion can help you measure the change in speed during the stroke cycle and optimize your technique based on it: https://www.rowinginmotion.com/optimize-your-rowing-stroke-with-rowing-in-motion/

      I’d say though that this is totally fair though. Since all athletes start at the same time and row the same time, they could (very theoretically) adjust their strokerate in a way that they reach the finish line just before the catch of the next stroke (where the boat is fastest). In practice it’s a matter of luck, I agree. But everyone has the same chances, given their lanes are of equal length.

  • Ian Watson

    Are the boats aligned before the boots on the ACM are raised? Surely the margin of error is in the boats being loose in the boot? If we watch the footage of the Olympic regatta , boats move backwards to the steak boat and then forward into the boot. So as I see it the boots are aligned but not the boats in them. So a margin should possibly be given the other end in the case of a close finish. Something akin to a side by side free start is deemed to be a dead heat if the margin is less than 3 feet. Can the alignment judge see through the ACM plastic boots to see if the boats are level.

    • Johannes

      One would hope that the volunteers holding the sterns of the boat gently push the boats into the boot. When I was actively racing, they always did.

  • Ian Watson

    Remember these being first introduced at a regatta we raced in 1994 in Paris. (Winning Lwt 4- in 5.48.86). I think it was Niall o’toole who managed to hit the boot and the stern lifted up as the boot went down!! Technology may have changed but this scenario would suggest there is a lot of room for movement if Niall could push into the boot early. I just took a still photo of the TV image just before the start of the Olympic singles and Aggreing with Kevin Armstrong I don’t think the boats are level in the boots.

  • Chris A

    The fact that there was an ACM is completely immaterial. The boats were still aligned by the Judge at the Start and aligner to the same start line as defined in the rules. Provided they are placed to allow this to happen their exact positioning doesn’t matter.

  • Peter

    In adition it would be better to make three pictures for each pair of boats. By taking tree picture you can approximate speed and accelaration of both. so you have to make on picture in front of the lane, second on lane an last when crossing the lane. After all, the precision is yust good as the setup and in my opion it can’t be so good. For example a half degree of the spoting lane caused round about 7 cm by a distance of 8 m between the lanes of boats (tan0.5degree * 8 m). So every race you have to make a new setup and control this again and again. no measurement technician would follow this precision after all. But for FISA rules the lenght isn’t so important, just the crossing lane.

  • Derek

    This seems easily solvable by simply having a second camera at the start. This would allow the actual difference in bow ball position to be calculated. Each boat could be determined to be x/1000ths Behring the leading bow ball and these margins could be subtracted or added after the boats finish. While doing this math might mean that boats might actually only have raced 1999.5784 meters, it would provide exact differentials from start camera to finish camera and be much more accurate.

  • john w

    The ACM holds the boats straight, there is still an aligned and Judge at the Start. The aligned at the Olympics and at World Championships is looking at video displays of the start line, same equipment that is used at the finish. The boat holders still move the boats in and out of the clog until correct alignment is certified by the judge at the start.
    When the stater presses the start button, the video freezes to show fair starts.

    • St.

      That’s only partly true. The equipment at the aligners hut and the finish tower differ in the way the cameras operate. The finish camera takes only one pixel column (or line), but several per second. The photo finish image is generated by stacking the 1px-column “pictures” horizontally. So you have time on the horizontal axis and a spatial “one-dimensional” picture on the vertical axis that build the photo finish image. At the aligners hut, it is just a regular two-dimensional i.e. spatial image at a specific point in time, i.e. when the start signal is triggered.

    • Ian Watson

      That clears a lot of errors up. Thank you. But if the times cannot be split should the result? If we cannot determine time to the slightest fraction in real time should we then revert to other means or as in swimming award the same result same colour medal.

  • Ken Shirriff

    The article says that 3cm is about the distance a swimmer travels in 1/1000th of a second. There must be a typo there, since that speed works out to 67 miles per hour. A more realistic number would be 2 or 3 millimeters.

  • Mario Haner

    How can the superimposed finish line be exactly vertical, when it is not in the middle of the viewport? It seems to me that if it’s off the center of the viewport, you’re not looking at the track in a 90 degree angle, and therefore there should be some perspective skew on the superimposed finish line?

  • Mette Bloch

    I agree!
    As the alignment can in now way be that precise! – and therefor you can NOT have a nerdy precicion in the other end of the lake! – no matter how har you try, the boxes are still in the water. Water is strong, and will move them a bit… just enough to judge someone a winner of a god medal, even though the “runner up” has excactly the same time. Rowing should NOT be messured down to 1/1000 – I even think that 1/100 is a bit unfair, as argumentet, the conditions down the course will always be unfair, no matter how flat the water.
    Water depth is different, currents are different, small ripples from motor boats etc.
    I was so dissapointed in the decision in this race. It IS a dead heat, end of story!

  • Dave P

    You’re missing the point: the rules say whichever boat crossed the finishing line first, not whichever boat completed the course in the shortest time. While it might seem like splitting hairs, the ‘time’ argument is subject to the granularity (correctly) alluded to above; whereas the spatial argument is subject to the resolution of the camera which is at least one order of magnitude greater.
    Additionally, an extrapolation can be made by means of a high frame-rate camera (cellphones do this just fine these days). Take two frames 1/100th a second apart with the leading boat either side of the finish line. If this shows the same boat ahead in both frames it’s safe to say that it was ahead at the finish line.
    Either which way it’s tiny compared to the fairness problem currently faced by swimming 🙁

    • Johannes

      The point about space vs. time is moot in my opinion. It’s discussing the same things at the end of the of day. I’m saying it is unfair to not declare a dead-heat when the boats are separated by 2cm at the finish if you can’t guarantee they weren’t separated by a smaller amount at the start. It’s the same when applying boat speed to this equation and saying 2cm at 5m/s is 4/1000 of a second (4ms).

  • Jeff

    I don’t remember there being mention of the position of the finish camera being precisely square to the course. It would seem that even the tiniest offset from the finish line would show up as more than a cm across six lanes. I’m new to rowing, but have been involved in track for many years. Is the current system in use anything like FinishLynx, where it s-t-r-e-t-c-h-e-s out the video to magnify the finish? Obviously, alignment issues make a difference, regardless. And how about courses with a turn? Lake Natoma comes to mind-there’s a big, right-hand dogleg in that course. The inside boat travels less than the outside boat. Should there be a stagger? Again-I’m new to this…just spitballing.

  • William Donegan

    Excellent article raising many good and valid points. It is a fact that the degree of accuracy used to determine the order of finish is not applied at the start. As “john w” points out, there is a freeze frame picture from a video camera which appears on a computer monitor which shows the position of the bows when the start occurs. The Judge at the Start should stop the race and declare a false start if the camera picture shows a bow across the start line. There is no specific requirement in the Rules of Racing for a camera at the start which has the accuracy used for the finish (at least 1/100th second). In practice, the start camera usually does not have the degree of accuracy (although I do not know what type of camera was used in Rio) to detect a bow which may be 2 cm over the line at the start. Even if it did have that accuracy, it can often take a few minutes while the photo is manipulated to finally determine the matter. By that time the crews will be well down the course, and, stopping the race will not be a reasonable option that would be acceptable to the rowers and coaches. In addition, it is almost impossible to get a perfect alignment on the start line in conditions where there are wind and waves. However, the alignment system used at the Olympics does increase the accuracy of the start, but not to the degree that it will always provide a perfect alignment on the start line. The “clogs” do move with the wind and waves.

    My opinion on the issue is that rules should be amended to allow that if the margin at the finish is less than a specified amount then the race is declared a dead heat.

Leave a Reply

Your email address will not be published. Required fields are marked *