Tuesday, 15 August 2017

Spider/Radar Charts added to Golf Predictor

Jordan Spieth's average finish in multiple historical metrics ahead of the 2017 US PGA Championship. Shows him relatively strong in his last five events and relatively weak in similar weather.

Comparing the field rankings for multiple golfers (tee-off group) ahead of the 2017 US PGA Championship. Love in particular doesn't have FR data for many metrics, so he is assigned the large default value for those.

As above, but with Padraig Harrington's data highlighted for extra legibility.

I am pleased to announce the addition of useful new spider (also known as radar and less commonly, star) charts to Golf Predictor. Spider charts are a type of statistical chart that show the difference in multiple parameters visually at a glance and as such, are very useful for comparing certain stats on Golf Predictor. This can be seen in the screenshots above, for the 2017 US PGA Championship. The main components of a spider chart are:
  1. The spokes emanating from the centre of the chart. The spokes are displayed equidistant from each other and there is one spoke per parameter displayed.
  2. The data points are shown on the spokes. The further away the point is from the origin at the centre, the bigger the value is. On Golf Predictor, this means the worse the stat is.
  3. The points (the parameter values that make up the data set) are joined together to form a polygon. The number of sides to the polygon and its shape depends on the number of parameters displayed on the chart and their values.
  4. There may be more than one data set (polygon) displayed on the same chart, e.g. the data for more than one golfer may be displayed with different colour polygons. This is shown in the second screenshot above, where the polygons for the golfers in a tee-off group for the 2017 US PGA Championship are plotted in different colours.
  5. The concentric grid lines on the spider chart are for scale purposes. It is possible to number this scale, but it doesn't look good and was somewhat confusing with the charting library (see below) that was used.

While ideally, a golfer should have the best statistics possible, in the real world, most top golfers will have some mixture of relatively good, so-so and bad stats. A spider chart tells you immediately where a golfer's relative strength and weaknesses are. For example, the first screenshot above, you can see the average finishing position for Jordan Spieth in multiple relevant historical categories. This shows that his recent performances (last five events) have been relatively good as this value is close to the origin. This, of course, is thanks mainly to his win in The Open! You can also see at a glance that his results in similar weather to what was forecasted for Quail Hollow are relatively bad, being furthest from the origin of the chart.

The ideal spider chart for golfer performance (where small values are good) is a polygon with all data points being as close to the origin as possible. Spieth's chart above is not that far off this, except for his similar weather performance. Some notes on these new spider charts:
  1. There has to be at least three data points to draw a spider/radar chart (you can not draw a polygon without at least three points!).
  2. Initially, I wasted a lot of time trying to use the Flot spider charts plug-in (Flot is used for all the other charts on site, bar the box plots), but I was not happy with the features or the documentation. Eventually I found RGraph, which has some attractive and easy to use spider charts, so I switched to them instead.
  3. Zero data points (i.e. missing values) are left off the single polygon spider (i.e. single golfer) charts. This proved to be the best way to handle this situation. The number of omitted zero point categories is printed above the chart. Leaving them in caused readability issues for certain data sets, as did defaulting them to a high number, e.g. 250.
  4. Zero data points are included on the multiple polygon (i.e. multiple golfers) spider charts, as some of the golfers charted may have a value for that parameter. In this situation, the best option was to have the missing values default to a value slightly above the highest value plotted. When this happens, a message is printed above the chart with the default value. 
  5. Despite the above point, some of the polygons (i.e. data for some golfers) on a multiple polygon spider chart may be difficult to read, depending on the values for the golfers chosen. For example, if you chose to compare stats for someone like Rory McIlroy and someone with poor or no stats, McIlroy's polygon will be hard to read, due to having to plot very good and very bad stats on the same chart. The only workarounds for this are to omit the golfer(s) with poor/no stats from the comparison or compare the stats for similar level golfers (this obviously is not an option if you are comparing a tee-off group!).
  6. Similar to above, a polygon on a spider chart with one or two bad values may be hard to read if all other values are good values. For example, the field ranks stats for Jordan Spieth ahead of the 2017 US PGA Championship are unbelievable, except for his tournament history rank (and his cut streak rank to a much lesser extent). This makes his field rank chart somewhat difficult to read.
  7. On multiple polygon (i.e. multiple golfers) spider charts, you can click on the golfer names in the legend (key) to highlight the values for that golfer. That is very useful for charts with data for more than one golfer, especially where the data overlaps considerably. This is shown in the third screenshot above, where the polygon for Padraig Harrington in the tee-off group for the 2017 US PGA Championship has been highlighted.
  8. The number of polygons (i.e. golfers) on a spider chart is limited to a maximum of four for legibility reasons. However, any more than two can be difficult to read,  but as mentioned above, you can highlight a polygon to see the data for a golfer more clearly.
  9. It was not possible to show the data point values on the chart. However, you can hover over a point to see its value, as shown in the first two screenshots above.
  10. The chart scales automatically, based on the biggest value to display. They may mean that a spider chart for a golfer with all good values may look similar to a chart for a golfer with all bad values. Clicking on a data point will indicate whether the values or all good or bad.
  11. Some of the new spider charts have average values displayed above the chart, where this is useful, e.g. the overall average on the early/mid/late season chart (see below).
  12. I have added a new icon for spider charts. At the very small size (14x14 pixels), it was difficult to create one that didn't just look like a blob! 
  13. You may have to reload any page you have opened in the recent past in order to see these new chart links.
  14. The overview table on the Compare Groups page had slightly different columns from its counterpart in the Compare Golfer page. That has been corrected and the same region performance has been added to the table on the Compare Groups page instead of the world ranking (this is available by choosing the last option in the 'Stats to Compare' dropdown list.
There has been a total of six new spider charts added to the pages on the site that show statistical information for a golfer or set of golfers. Specifically, the spider charts have been added to the following pages:
  1. Prediction Data page: two single polygon (i.e. single golfer) spider charts; one for field ranks and one for historical performance (e.g. same course, season, last five events etc.) for each golfer in the field. These two new charts are accessible from the General panel.
  2. Golfer Data page: one single polygon spider chart for regular/major/WGC/FedEx average result in each golfer's career.
  3. Season Segment Data page: one single polygon spider chart for early/mid/late season performance for each golfer.
  4. Compare Golfer/Compare Groups page: one multiple polygon (i.e. multiple golfers) spider chart for the historical performance (e.g. same course, season, last five events etc.) of the chosen golfers/tee-off group.
  5.  Compare Field Ranks/Compare Groups page: one multiple polygon spider chart for the field ranks of the chosen golfers/tee-off group (see last two screenshots above).
This brings the total number of charts on the site to 485. I trust you will find these new spider charts useful as a graphical representation of (multiple) golfer performance in certain key metrics. Just another way to make Golf Predictor even better!


    Monday, 14 August 2017

    2017 - Week 33 Predictions/Statistics Online

    The predictions and statistics for this week, the Saltire Energy Paul Lawrie Match Play (European PGA Tour), the Fiji International (European PGA Tour) and the Wyndham Championship (US PGA Tour), are now available on Golf Predictor. While there aren't many big golfing names teeing it up in either European Tour event (especially in Fiji!), there is a solid list of European Tour stalwarts teeing it up at the relocated Paul Lawrie match play event. Although Golf Predictor is designed primarily for stroke play tournaments, the match play event in Germany has also been put through the prediction engine. Note that the course in Fiji has been significantly updated since last year by local hero Vijay Singh. There are not many big names competing in America either this week, with the Wyndham Championship coming straight after the last major of the season However, there is a host of middle tier guys trying to improve/maintain their place in the FedEx Cup standings ahead of the upcoming play-offs..

    2017 - Week 32 Winner

    Justin Thomas (ranked 32nd by Golf Predictor) won the US PGA Championship on the European/US PGA Tour. The 24 year old American secured his first major title with an impressive final day performance at Quail Hollow. When most faltered after the turn on Sunday, Thomas came into his own and won by a comfortable two shots, even with a closing bogey. Our top ranked player, Hideki Matsuyama, finished poorly on the back nine for tied 6th. Overall, we had ten of the top thirteen plus ties (10/21) on the tough layout.

    Tuesday, 8 August 2017

    2016 - Top 5 Hardest Courses/Tournaments on the European PGA Tour

    Based on the average score in relation to par in the tournaments hosted on them, these were the 5 toughest courses on the European PGA Tour in the 2015 GP Season (calendar year):
    1. Club de Golf Valderrama (Real Club Valderrama Open de Espana)
    2. Oakmont CC (US Open)
    3. Four Seasons GC at Anahita (AfrAsia Bank Mauritius Open)
    4. Augusta National GC (The Masters)
    5. Royal Golf Dar Es Salam (Trophee Hassan II)
    Granted, the weather and relatively weak fields may have played a part in some of these events, but on the other hand, the list is similar to lists from previous years. You can get more detailed information by looking at the 'Highest Average Round vs Par' chart on the 'Other Stats/Tourn. Stats' page in the member section of Golf Predictor.

    Click on the "Lists" category opposite to see more such lists.

    2016 - Top 5 Easiest Courses/Tournaments on the European PGA Tour

    Based on the average score in relation to par in the tournaments hosted on them, these were the 5 easiest courses on the European PGA Tour in the 2015 GP Season (calendar year):
    1. Oceânico Victoria Clube de Golfe (Portugal Masters)
    2. Earth Course, Jumeirah Golf Estates (DP World Tour Championship, Dubai)
    3. Golf Resort Bad Griesbach (Porsche European Open)
    4. GC Milano (Italian Open)  
    5. Regnum Carya Golf & Spa Resort (Turkish Airlines Open)
    This is somewhat similar to the same list in previous seasons. You can get more detailed information by looking at the 'Lowest Average Round vs Par' chart on the 'Other Stats/Tourn. Stats' page in the member section of Golf Predictor.

    Click on the "Lists" category opposite to see more such lists.

    2016 - Top 5 Hardest Courses/Tournaments on the US PGA Tour

    Based on the average score in relation to par in the tournaments hosted on them, these were the toughest courses/tournaments on the US PGA Tour in the 2015 GP Season (calendar year):
    1. Oakmont CC (US Open)
    2. Augusta National GC (The Masters)
    3. Royal Troon (The Open Championship)
    4. PGA National - Champion Course (The Honda Classic)
    5. Innisbrook Resort & GC (Valspar Championship)
    As usual, the US Open is the hardest tournament, but it's joined at the top by two of its fellow majors. You can get more detailed information by looking at the 'Highest Average Round vs Par' chart on the 'Other Stats/Tourn. Stats' page in the member section of Golf Predictor.

    Click on the "Lists" category opposite to see more such lists.

    2016 - Top 5 Easiest Courses/Tournaments on the US PGA Tour

    Based on the average score in relation to par in the tournaments hosted on them, these were the 5 easiest courses/tournaments on the US PGA Tour in the 2016 GP Season (calendar year):
    1. The Plantation Course (Hyundai Tournament of Champions)
    2. PGA West and sister courses (Humana Challenge in pship with the Clinton Foundation)
    3. Kuala Lumpur G & CC (CIMB Classic)
    4. RTJ Golf Trail (Grand National) (Barbasol Championship)
    5. Waialae CC (Sony Open in Hawaii)
    The top three retain their positions in the list from the previous year and Hawaii is obviously the place to go for a birdie-fest!! You can get more detailed information by looking at the 'Lowest Average Round vs Par' chart on the 'Other Stats/Tourn. Stats' page in the member section of Golf Predictor.

    Click on the "Lists" category opposite to see more such lists.

    Monday, 7 August 2017

    2017 - Week 32 Predictions/Statistics Online

    The predictions and statistics for this week, the US PGA Championship (European/US PGA Tour), are now available on Golf Predictor. It's the final major of the season and the PGA Championship heads to Quail Hollow for the first time. It's the last chance for eight months for a golfer to land one of those prestigious titles, so competition is bound to be fierce!

    2017 - Week 31 Winners

    Hideki Matsuyama (ranked 4th by Golf Predictor) won the WGC - Bridgestone Invitational on the European/US PGA Tour. The 25 year old Japanese star won his second WGC title with a tour de force in Ohio. Matsuyama put on an exhibition of ball striking on the final day to set a new course record of 61 (-9). In doing so, he blew away the field and won by five shots. Our top ranked player, Jordan Spieth, finished in tied 13th and we had twelve of the top seventeen plus ties (12/23) in total.

    Chris Stroud (ranked 64th by Golf Predictor) won the Barracuda Championship on the US PGA Tour. The 35 year old American secured his first tour title at the 290th time of asking with a play-off  win in the modified Stableford event. After Stroud finished with a five point eagle to post 44 points, he was matched by Greg Owen and Richie Werenski. However Stroud wasn't to be denied his first win and he prevailed at the second extra hole over Werenski. Our top ranked player, Daniel Summerhays, finished in tied 27th and we had only eight of the top eighteen plus ties (8/25*) in the second tier event. 


    *Due to non-finishes, Barber and Flores are promoted to the top 25 of the GP rankings

    Monday, 31 July 2017

    2017 - Week 31 Predictions/Statistics Online

    The predictions and statistics for this week, the WGC - Bridgestone Invitational (European/US PGA Tour) and the Barracuda Championship (US PGA Tour), are now available on Golf Predictor. The WGC event has been moved back to its regular slot after being moved to facilitate the Olympics last year. Fences seem to have been mended as it is once again sanctioned on the European Tour and boasting its usual strong field. Meanwhile, the second string Stableford event in Reno has to make do with only one golfer from the top one hundred in the world rankings as all the big names playing this week are teeing it up in Ohio.