Quantitative empirical properties of TagPro maps, two years later
With the recent discussion on TagPro maps, I figured it might be useful to redo my quantative analysis of empirical map properties from two years ago with today’s data. To recap, this analysis computes and visualises three indicators for each map:
- non-hold per grab, the average time it takes a team to grab;
- hold per drop, the average time it takes a flag carrier to die;
- captures per grab, the offensive effectiveness.
Mainly the first two define the style of the map, in a way closely related to the map archetypes that have long been used for categorising maps. A difference is that the archetype definitions also consider getting out of enemy base rather than just grabbing, but the indicators turn out to be quantitatively consistent with the archetypical examples nevertheless.
So, I took yesterday’s bulk data from the TagPro Analytics database, filtered public matches with three or more starters on each team (to avoid distortion by waiting) and plotted the maps with at least one hundred such matches: click for larger/interactive
Some random observations and thoughts compared to two years ago:
- The analysis method still works well. For maps that existed two years ago, the differences in relative positioning are small. Differences may be attributed to the higher quality and homogeneity of the current dataset, in addition to shifts in general game playing style.
- A line of neutral-flag maps has now emerged. Neutral-flag is literally a one-dimensional game mode because one team’s hold is basically the other team’s non-hold. For the same reason, the game mode is relatively defensive.
- While capture-the-flag is a two-dimensional game mode, many of the newer maps lie close to the traditional hard-to-survive Wormy-Star line. While currently played maps still show large but seemingly declining variety in non-hold per grab, the hold per drop has become much more similar than years ago. In that way, the game is indeed getting stale. On the other hand, Mode 7, Dealer and Atomic have recently pushed the balanced/fast-paced frontier significantly further, so we still see innovation.
- The rise of neutral-flag as a game mode aligns with and perhaps leads the development within the capture-the-flag mode: towards flag carriers that die quicker.
- Based on the indicators alone it is hard to explain map ratings. Except for the Event Horizons, maps that are hated often appear reasonably similar to maps that are loved. The popularity of Scorpio is surprising considering its extreme positioning and the number of disliked maps nearby.
- While Smirk and IRON are very much liked, no recent maps have indicators even close. There is also a large gap between them and GeoKoala, which has been popular for a long time. Boombox and Hornswoggle are somewhat nearby and also have very positive ratings. Perhaps map makers and the map test committee need to look into hard-to-grab maps more: making (re)grabs harder could be an attractive alternative solution to the ‘regrab problem’, as opposed to making returns easier.
Want to do an analysis like this yourself on TagPro Analytics data? Download the C++ source code that generated this graph and compile it against the TagPro Analytics C++ headers.