Time Trial Rankings

How are the points for each ride calculated?

Points are assigned by assessing a rider's time in an event and comparing it with the times of other riders…

Points are assigned by assessing a rider's time in an event and comparing it with the times of other riders who finished in a similar time. By assessing the form (i.e. current ranking) of those riders it is possible to make a numerical assessment of each rider's achievement.

In essence the points a rider gets from an event are a notional position the rider would have achieved if everyone currently riding in open events had entered this event.

As it is a positional system, the better the ride the fewer points a rider gets.

It makes no difference what the makeup of the field is as each rider is awarded points relative to the abilities of those who finish near them, so riding in a specialist event (such as juniors, ladies, vets or even road bike only) should give similar points as a rider would get in an event open to all. Similarly the speed or length of the course should have no bearing on the results as they are simply calibrated by your fellow competitors.

The way this is done is by plotting a graph of time against known ranking, ensuring a best fit but taking variations in the field into account. Once you have a suitable curve for each event you can then see what each rider`s actual time would be worth in ranking points. The more riders who enter who already have a known rank the more accurate the results will be. In the body of the curve the points will be very accurate but significant outliers are more difficult to define and futher modelling is used to give them the most appropriate points.

Over time with more and more data available the acurracy should improve even more.

Posted 2 years agoby Roger Clarke

Why do some rides not have points allocated?

If an event has a very small number of riders it is not possible to reliably deduce points for the…

If an event has a very small number of riders it is not possible to reliably deduce points for the event as there isn’t enough data to generate a valid result.

Generally there needs to be at least 10 riders for any sense of validity. Some events may have more riders than this but if only a few of them already have a known rank then this prevents the points from being generated accurately.

If an event takes place on the same day and on the same course and over the same distance as another event the two will be merged for the purpose of point scoring (although this won't show on the website). This enables events with low numbers to be counted if at all possible.

Posted 2 years agoby Roger Clarke

How does the ranking get generated?

The overall ranking is supposed to reflect a rider's current performance level in respect of their competitors (i.e. their current…

The overall ranking is supposed to reflect a rider's current performance level in respect of their competitors (i.e. their current form).

To do this the more recent races available are used to generate an average score. The more results a rider has and the more recent they are, the more accurate a rider's rank will be. The very best and worst of a rider's results are excluded where they can.

A rider needs 6 races in the last 12 months to be complete and show in the main ranking table. However, you can filter by all results to show provisional rankings for everyone irrespective of how many rides they've had although these will not be as accurate.

There are three basic scoring situations to decide on which are a rider's qualifying races:

  1. If a rider has more than 8 races in the last 4 months then their overall score will be the average of the six best of these (excluding the very best one).
  2. If a rider doesn’t have this many recent races they will get the average of up to 6 of their most recent races (up to 1 year old) excluding the best and worst rides.
  3. If a rider has fewer than 4 races in the last 12 months they will get an overall score which is the average of all of their race scores.

The reason for these distinctions is that ideally a rider's current form should be based on their best recent results. If this is not possible then it should be based on their most recent rides, accepting that some of them will be older; and if you have very few (<4) the best and worst rides must be kept as, even though they may not be accurate, it is better than relying on even fewer results.

In creating scores and ranks an attempt has been made to put all riders in order of ability. This is not an exact science as it could be argued that you are only as good as your last race or conversly that a full 12 months rides gives a better reflection of ability. Clearly when riders are seperated by just 1 point over a range of 10,000 it is unlikely we can be sure one rider is definately stronger than the other. As such the categories are probably more realistic. However the detailed breakdown is available so you may as well enjoy it.

See "Can the system be more accurate" for more details about other ways to rank.

Posted 2 years agoby Roger Clarke

Why are the best and worst results not counted?

Once a rider has enough results (4 or more) their best and worst results will not count towards their overall…

Once a rider has enough results (4 or more) their best and worst results will not count towards their overall score. This is because it is possible to get aberrations in the results which may have too dramatic an effect on a rider's ranking.

For example, a rider may puncture or fall and still finish but in a time far outside what would be expected of them, or just have a bad day, not representative of their true ability. This can seriously affect their ranking for some time afterwards if it is not weeded out. If it happens more than once the rider will have to accept the consequences!

Riders can get results which appear too good. If the weather changes during an event riders may end up beating others they would never expect to beat and get a really good score which would not be representative of their ability and would inappropriately improve their ranking.

Also, at the highest levels of competition, if a rider is miles ahead of anyone else it is very hard to truly evaluate the merit of the ride and the algorithm sometimes gives an inappropriately good score, or if a very well seeded rider doesn’t perform as expected the system can sometimes inflate the scores of others who beat them.

This may seem hard but a rider's ranking should be a reflection of their general recent results not significantly influenced by one extreme result, and this shouldn’t make much difference if a rider is fairly consistent in their performance.

Posted 2 years agoby Roger Clarke

What is a provisional ranking?

A rider needs to have ridden 6 events in the last 12 months to have a “qualified” ranking status. Before…

A rider needs to have ridden 6 events in the last 12 months to have a “qualified” ranking status. Before this a rider will have a "provisional" status.

There is an option in the rankings list to include provisional riders in the list or not.

Provisional riders will be shown with the rank they would have should they have a "qualified" ranking status with the same score.

If a rider's most recent ride is more than 12 months ago they will be given a status of "lapsed" and will not appear in the rankings list.

On recording another result they will receive a "provisional" ranking status again based on the new result alone.

Posted 2 years agoby Roger Clarke

What is the difference between points, score and rank?

Points are what riders are given for each race they ride in. The fewer points the better the rider performed…

Points are what riders are given for each race they ride in. The fewer points the better the rider performed in that race.

The score is the average of the points in the counting rides. See "How does the ranking get generated"

The rank is the rider's position amongst riders with at least 6 rides within the last year. This is simply generated by sorting the riders by ascending score. 

When the rankings list is filtered by gender the rank displayed is the position for a rider within riders of that gender. If a list includes riders of both genders the rank displayed is the position for a rider within all riders, or "overall rank".

When the rankings list is filtered by other criteria such as district, the ranking shown is dependent on whether the list is also filtered by gender or not as above. In addition the position within the set of filters chosen is displayed.

Posted 2 years agoby Roger Clarke

What events are used to generate the points?

Counting events must be solo time trial events. No team time trials are included. Tricycles and tandems are also excluded. No…

Counting events must be solo time trial events. No team time trials are included. Tricycles and tandems are also excluded. No hill climbs are taken into consideration either.

Times over 7 hours are excluded, unless they are specified distance events (ie 12 or 24 hr races)

If events have an unusual status e.g. road bike, they will be taken in isolation so the comparison with other riders is fair, but if road bike categories are included in an event open to all then they are assessed along with other riders on different machines and the points are likely to be worse for those riders because of this.

Events taking place during November, December or January do not have points awarded as these events are not considered to necessarily reflect a rider's form. 

See also "Why do some rides not have points allocated."

Posted 2 years agoby Roger Clarke

How do rider categories work?

The category is simply a system of ordering riders by rank and giving them a percentile from 1 to 100…

The category is simply a system of ordering riders by rank and giving them a percentile from 1 to 100 across the whole range of riders.

A percentile is not just a simple division into equal 100ths of the field, but is weighted so the middle (much more common centiles) have far more riders in than the outliers. This is based on a normal distribution (for those interested in such things). As such there will be far fewer riders who have a category of 1 than have a category of 50.

The percentiles have been adjusted into 5 bands of 1-20 and given a letter, so a percentile of 1 will give you a category of A1. This runs up to A20 and then B1 takes over. Category 50 would be C10. As such the bands are A,B,C,D and E, with most riders in the middle bands. You could equate this to BC categories Elite to Cat 4 if you wish. A1 is the very best band and currently has fewer than 10 riders. However our categories cover a much wider ability range - arguably most "Cat 3" BC cyclists would be in the C or even B bands for time trials. You`ll be able to judge this for yourselves by riders you know who do both.

You recieve a new category after every ride, depending on your ranking position. This shouldn`t change very much once you are an established rider but can flip between 2 letters if you are on the border between categories, eg between D1 and C20.

Categories show in a couple of places on the site -

The rider screen shows the current category top right but also the category after each event on the events listing for that rider so you can see how a rider is progressing over time.

The individual events screen shows the categories of each rider who has a category before the event so you can see how well they did compared with other rider`s by category and see who has "won" in each category.

Posted 2 years agoby Roger Clarke

Can I refine the data to see local riders or other groups I compete with

You can filter the rankings in various ways to get more specific information. You can include provisional riders or filter…

You can filter the rankings in various ways to get more specific information.

You can include provisional riders or filter by gender or age ranges.

You can filter by club or district.

In this way local areas can have local ranking competitions.

It may be possible to make more specific selection available over time.

Posted 2 years agoby Roger Clarke

What do all the up and down arrows mean?

There are arrows at the side of the Rankings, Event and Rider individual pages. Rankings list Up and down arrows…

There are arrows at the side of the Rankings, Event and Rider individual pages.

Rankings list

Up and down arrows on the rankings page show the change in rank in the last week (assuming a weekly update). It is possible to move rank without riding as other riders move around and some of a rider's old races lapse from the timeframe for scoring.

A dot shows a new entry into the completed ranking and a horizontal arrow shows a non-mover.

Movements in ranking are accurate for the overall rank and for gender ranks but do not apply to other filters.

Event page

On the event page in the list of results the arrow shows in the position column. This is a marker of the change in position from what would have been predicted from a rider's rank before the race began.

So if a rider has an up arrow and a 5 next to it it signifies they have finished 5 places above the position they were predicted by their rank before the race.Riders with dots here did not have a rank before the event and so are not taken into consideration when predicting the positions.

As such a ranked rider will not be penalised in position change by a rider who doesnt have a rank beating them or vice versa.

Due to this it is possible to track back changes in position and find that two riders must have been predicted to finish in the same place, but this is an anomily of new riders places not being counted. 

Rider page

On the rider page the arrows against the points for each event signify the change that has taken place in the score from this event.

For example, if you have a score of 1000 and the event gives you 900 points then you would see an up arrow as you have improved your score.

This is not always intuitive as it is possible to have points from a race which are worse than the points from the last race but still have an up arrow if your score (compiled from the last few events) becomes better than the latest race results points. If your last few races have up arrows you can be sure you are steadily moving up the ranks even though there are variations in each of your rides.

Also remember your score is not the same as your overall rank. (Score is most closely aligned to your overall rank including provisional riders - currently up to about 10,000).

Posted 2 years agoby Nick Wild

What do the buttons on left of the rankings list show?

These buttons reveal the recent races for the rider and show which ones have points which count towards the rider's…

These buttons reveal the recent races for the rider and show which ones have points which count towards the rider's score (highlighted)

See more about how these are chosen in "How does the ranking get generated".

Posted 2 years agoby Nick Wild

Can you give me more detail on the way points are allocated?

For those who would like a deeper understanding of the way points are allocated here goes! Here is a typical…

For those who would like a deeper understanding of the way points are allocated here goes!

Here is a typical graph for a race. This one is the J2/3 10 from Manchester on 25th March 17. The x axis is the time in seconds for each rider who completed the event and had a rank before the race began. The Y is their "pre-rank". This gives us the plot points, each one representing a rider`s ranking and time. Using this data it is possible to fit a curve (blue line) which reflects the relationship between time and rank for the event. You can then map each ride onto the line to give an approximation of the ranking value (points) for the event itself.

You can see that the line in the middle is fairly straight suggesting a linear relationship between ranking and times, but curves at the ends for outliers. One of the problems is that the outliers have, self evidently, the fewest other riders aroud them to validate their position, and because they tend to be significantly in front of the rest of the field the curve becomes unreliable and can go negative. At the other end of the spectum the same applies to slower riders whose rankings do not fit closely with their times and can cause a dip which makes no sense (a slower time getting a better rank).

For this reason it is necessary to ajust the end points to give the red line which reliably increases from one rider to the next and is matched for the highest perfoming riders - more on this below.

Same situation again, this time its the Nelson Whelers L503 form 2nd April 2017. Exactly the same principle applies and the curve is a similar shape for a different distance.

Here's a 25 with a lot of fast riders entered. Its the Lea Valley 25 from 8.4.17. The black line is the first fit line. This includes all pre ranked points but you can see some riders are a long way from this i.e. their pre rank doesn't make sense compared to their perfomance on the day. This tends to be a problem with earlier races in the 2 year data set as the ranking has not had a chance to settle down. These riders are excluded before the red line is set which explains why it is lower between 3000 to 3200 minutes as the erroneous pre ranks had pulled it too high. You will be able to work out which riders these are by seeing big plusses for their position improvement in the race results. Next time they ride their rank will be closer to the curve if they perform as expected.

This race also shows the left hand flat tail when lots of fast riders enter, and the extremely fast rides by a small handful of riders. This is the challenging area to get the ranking right. As you can see the main body of the graph is still a straight line.

Heres the V718 from 17th April 2017. You can see the cluster of riders at the top end of the field but again a similar shape but now with a bigger "flick" at the top end.

One final graph just to show another regular senario.

 

At first glance this looks a diffrent shape. Its the VTTA east anglia E33/25 from 15th April 17. See how closely the pre ranks are starting to fit the curve, just 2 months into the system. The flat top compresses the earlier rides but actually the majority of the riders are still in the straight line area. You can clearly see here how the rank can go negative.

So, how to mange the top end. Have a look at the next graph.

This is a graph of 2 years of everyones pb at a 25 in time order. Clearly there's a pattern here.

And here's the top 300 rides. The proposition is that we can use this nice smooth prediction to reference one ride to another, so for example if you know the time someone rode and their rank, and you know someone else`s time you can predict their rank. This is the premise for setting the points for the top end riders based on rides further down the field and avoiding the variance from the original red line graphs.

But take a look at this.

This is everyone 25s for 2 years based on their rank on the day. Clearly there is a mid line in there somewhere but the definition has gone. 

Here's the top end of this graph.

 

If we can find a line in here we can use it as suggested to set other ranks at the top end. Its not as neat but here we go. Last graph I promise.

You can see the line goes to pot at about rank 1000 but is fairly stable till then. I have developed tables using this data to compare fast rides on the day with other rides lower down the field so we can fairly accurately set a score to a ride even if its hundreds or thousands of points above. All the fastest rides in every event are validated in this way. Its not perfect but its reproducible.

I would like to say thereby we can give an exact score for any ride. This is fantasy however as no such data exists or ever could. Everyone rides differently on different days and exceeds their previous best or has a bad day or their ranking isnt a fair comparison because they`ve only ridden once and someone else is a frequent rider with 20 races data, or any host of other variations.

Finally a point about variance and the very best riders. 

I said we could think about the points as how everyone would finish in a race with all participants. So you should expect a top flight rider to finish very close to his overall rank most of the time. Well, this isn't really true, although its a helpful way of thinking about the points. 

The points need to reflect the quality of the rides, not the riders, so it would be possible to identify the best ride of the year and give this 1 point (remembering that this is only a concept as such a thing couldnt really exist - who would ever agree?) . However its not very likely that the number 1 rider's rides are exactly equally good so he gets 1 point every time. If you based everything on this endpoint you would find many riders who had fantastic results would still get their best rides 20,30 or 40 points down from this notional end point. Also its possible no one can get close to the very best rider most of the time so their points will reflect this. For this reason and because every day is different and the graphs are reflecting real life and not made up data, there is some significant variance so even riders in the top 20 will have point scores which change significantly from one race to the next. This should all average out and give reliable results over time.

 

Posted 2 years agoby Nick Wild

What do the little fires mean?

The fire symbols next to riders names indicate a rider who has "form". This means they are performing currently at…

The fire symbols next to riders names indicate a rider who has "form". This means they are performing currently at a higher level than their ranking suggests. Therefore its not necessarily the best riders who have form, but those who are rising up through the ranks.

This is helpful on predicted startsheets, because the predictions are based entirely on current rank, so a rider with form can be expected to beat their prediction.

To get form you need your points from a race to exceed your score by a certain percentage, consecutively on at least your last 2 rides, within the last 3 months. Statistically having form makes you about twice as likely to equal or beat your predicted position as not having form, and at any one time on average about 5% of riders will have 1 form fire and 5% will have 2. Two fires denotes a bigger percentage difference from your score and further enhances the likelihood of you beating your predictions.

The form shown on the startsheet is the pre-race form and that on the results is the post-race form.

Form has been populated retrospectively through 2017 and 2018 so you can look back and see the periods when you were improving the most.

Posted 2 years agoby Nick Wild

When and how are the events added?

Hopefully the system will be updated every Wednesday evening. This gives enough time for most of last weekends results and…

Hopefully the system will be updated every Wednesday evening. This gives enough time for most of last weekends results and next weekends startsheets to be in. The system takes a minimum of 30 mins to upload currently and if there are problems with the data potentially much longer. We are a small team (2) and are developing and running in our own time, but its rewarding and a lot of fun.

If results are late then they probably will have to wait until next week, as they affect the rankings and we cant afford to run the rankings more than once a week. However its relatively easy to run extra startsheets so if I`m aware a significant number of startsheets have become available I will try to fit them in later in the week.

Startsheets which are added for the weekend after next are pending; the predictions will be based on current rankings but they may be changed the following week in light of any new results which come in from the interveening weekend. 

Sometimes startsheets and results change on the CTT website. If a change is made it will probably be missed, as we cant keep checking back to see if there are any amendments. Spindata reflects the events at the time they were downloaded from the CTT website. Errors in times are a particular problem as the rankings are all relative so one processed its not really possible to back track them.

The startsheets and results are essentially independant however, so if you are not on the startsheet but ride the event you will get your points.

Posted 2 years agoby Nick Wild

Tell me about the startsheets

The startsheets can be seen for upcoming events, from the event pages. The first tab is the starting order as…

The startsheets can be seen for upcoming events, from the event pages.

The first tab is the starting order as listed on the CTT website.

The second tab is the predcited order at the finish, based on how long a rider is predicted to take on the course and what time they start. This gives riders and idea of who might come past them on the course and timekeepers an idea of who to expect and when.

The third tab, and the one which is the defualt entry for the page is the predicted finishing position.

The positions are generated based on the ranks of the riders so someone with a better rank is expected to beat someone with a worse rank. This can be stated for every course as its independant of the course type or length. Riders for whom we dont have enough information to rank will appear at the bottom of the list. This means your predicted position is likely be be affected by riders who are stronger than you but didnt have a rank before the event so you may well not attain your predicted position in the results, but the change in position column in the results takes this into account - ie you wont be penalised for being beaten by someone we couldnt predict.

The predicted times are based on previous events on this course and at this distance. If the course and distance havent been raced since the start of 2017 there will be no times. Sometimes events appear to be on the same course as before but there are changes to the course or there were roadworks or some problem last time and then this data cant be used to predict times. We`ve done our best to weed this out.

The predictions take no account of conditions. If a course has been run only once before the preditions will be based on this limited data and if that event was unrepresentative then the predictions are likely to be out, similarly if the current event has particularly good or bad conditions then the times will be wrong. Events which are run a lot will have more stable predictions, but they will be the average for the event so there can still be significant changes given the conditions on the day.

Riders with form (see "little fires") are more likely to exceed their predictions.

The predicitons will usually be in category order but this may not always be the case as frequent riders status can affect this.

Posted 2 years agoby Nick Wild