Posts
10
Joined
3/8/2009
Location
EL MIRAGE, AZ, USA
After seeing Sexton’s incredible gate timing at the muddy San Francisco SX, it got me thinking… how random are the gate drops?
“…at least 5 seconds, but not more than10 seconds,” according to the Rule Book.
As a geek, I did some analysis over 100+ gates drop in 14 SX rounds in ‘23 and ‘24. Feel free to look at the data and the analysis.
TL;DR:
It’s pretty predictable; just dump the clutch at 5.42 seconds after the board goes sideways in the your next 450 main.
I think we all know who to really ask....
Watch and listen for the sliding shaft..at the pivot point on the starting gate...
You're talking about the starting gate?
Well ya
The Shop
Luxon 4-Post Bar Mounts
$189.95 - $239.95
DeCal Works Huge Plastic Inventory of UFO and Polisport kits.
Better off staring next to the dog box, and start on movement inside the dog box, or if it's a pneumatic system listen for the valve...
Back in the day I would stare at a paint chip or a weld spot if it just quivered hammer down!
Back in the day, I'd wait for the rubber band to move!
I disagree with your simplification of 5.2 seconds though, looks to be a bi-modal distribution but I think your analysis assumes a normal distribution?
Also, what happens when the AMA violates their own rule and drops the gate <5 seconds after the card turn?
Looks like it happens close to 10% of the time.
Human error, a guy looking at a stopwatch activates the gate, would be better to push a button when the card turns, and have a random timer 5-10 seconds later activate the gate.
They penalize riders for human error, seems like a double standard
When I was racing in Germany in the early 70’s, the gates were forward falling mechanical contraptions of all types. One just went and studied the mechanism and then worked hard to qualify first in timed practice on Saturday. Line up where the mechanism was exposed and wait for a bit of motion.
In the late 70’s and early 80’s, there was a track in Yakima called Sun Target and it had a forward falling mechanical gate. One year at the Over the Hill Races over 6 starts (30 expert) I grabbed 5 hole-shots. Missed the 6th one to a friend who at dinner Saturday evening asked me WTF – I shared.
We all timed it poorly at the last race
Great point @3strokemx !
I do check for normalcy in pg 6 with the QQ plot:
If the data were exactly normally distributed, all points would fall on the straight line. I fudge it and say “close enough,” but it is a bit of a violation of the MLR assumption for the residual error (and by implication the dependent value being regressed ) to be normally distributed.
Perhaps a log-transformation gives a better fit…I’ll have to check it out.
Yeah, I always got good Starts at Sacramento Raceway as I would lineup where the doghouse was supposed to be but it was just a rope so as soon as I saw that rope yank, I popped the clutch just a little bit early and always got good starts lol
You’re supposed to look beyond the gate and have the gate itself in your peripheral
You're supposed to get the holeshot, by whatever method works for you!
Agree. IMO track which human is doing the gate drop, and group data per person. A truly randomly generated number would be the best way. This could be a mechanical system or a human using a random-number generator program. I can't recall ever seeing a gate drop after 9s.
I would be more interested in ability to predict timing on the gate drop for the main based on earlier gate drops that day. I have no data, but when I count from the stands it seems like basing it on the last few starts is most likely to be a repeat.
Pit Row
The 9-10 second mark feels real long both when you're watching, or racing.
The sooner the better before my heart jumps out of my chest......
Anyone here feel that?
I am surprised there are not more guys rolling before the start. If you are not a top 10 guy, why not. There is a bit of room for the bikes to be backed up a little. If you are rolling, even the slightest bit, you will smoke the guys that are completely still.
Your linear model appears to predict the most likely gate drop time from the dataset. Looking at your histogram though it's clearly bimodal at around 5.25 and 5.75 secs (not sure what your bin size was).
The problem is that predicting the drop has a binary outcome - you go sometime after the gate (success) or you go some time before gate (fail). The way the histogram looks that would have about a 50% success rate - no one is going risk that on an ongoing basis.
Perhaps on a case by case basis like SF where the difference between being out front and not was so great it was worth the risk.
I only know what's worked for me. And I'm not telling.
But I'm really enjoying the Statistics lesson.
It doesnt have to be a fail. If you "go smoothly", you can still pull in the clutch if you notice it ain't there.
They should expand to be between 5-7sec for something like 75-85% of the time, with occasional 8-9sec. The 10 I think you can just drop.
Johnny Moore?
You guys talking about listen to the gate are funny, I am sure you can hear a pin drop with 40 bikes pinned after the card turns sideways.
Post a reply to: Timing the Gate