Done : a)Fixed persistent display issue across resolutions b)Made the timer display during loading c)Made the timer pause during loading (it may count some milliseconds of loading but that's really small)

Currently, the timer always say "Run not started." even when it is impossible for a run to be started. I'll change it.

The config option in the UI is done.

This let the question of which times should be counted. I will disable it for loading times, but input is appreciated : are there other moments were it should/could be disabled ?

I have successfully moved the speedrun timer in new dedicated code files. The STK devs will probably find this much better than the hacky solution with the main loop.

I have also changed the look. The counter use the high res digit fonts used by the race timer ; making it look much nicer ; and I have moved it around to sit just to the left of the usual race time.

So 1 & 2 are done.

The timer's position is also relative now - no more issue when the game's resolution is changed.

The position of the timer is fixed. This plays badly with changing resolutions. Lots of things to change.

I updated the files for the latest STK dev version, and got the mod to work.

Now, it needs to be improved. The goal is to get it merged in the base game to reduce speedrun hassle for 0.9.4 and future versions.

1)It needs to look better. It's poorly placed and fuzzy. 2)It places its functions in the main loop file. I don't think this will be appreciated by the STK dev. 3)It needs to be enabled with an option (similar to the display FPS option). Otherwise, it won't be accepted. 4)It needs to not take into account loading times. 5)There may be other moments where the timer should not run : e.g. should it run during the initial cutscene ? This is related to the RTA-vs-IGT question. I think that not counting events is an issue if it creates situation when you can win IGT time by increasing RTA time.. E.g. ; if we counted the GP official time rather than the play time, spending two minutes to manipulate the issue of a race before skipping would be IGT-optimal but very bad for RTA. This is not good for the spirit of a speedrun imo. But I'm not so sure about it for things like cutscenes.

If I manage to do it, I'd probably backport the changes to make it skip load times for 0.9.3. I firmly believe that RTA without loading times (which is more or less IGT) is a much better measure of run quality than pure RTA.

I'm going through the code. It's great. But the way the timer works by counting load time defeats the purpose.

I've found a new shortcut in shifting sand. Showcased in this new IL WR run :

Don't gain much time for the first two laps, but more for the last one. In the last lap, it is probably worth it in the GP in full-game runs.

I think that the fact this has not been used until now is testament to how boring that track is (also, that it is a nitro challenge doesn't help ; it would be more useful if it was possible to use it twice in a run)

Something speedrunners will probably like : a new speedometer and nitro bar have been merged in the main STK code.

The speedometer can display a speed up to 400 against 305 previously (the numbers are not directly displayed, the quoted number are those from the old numerical speedometer) ; the nitro bar is segmented in 5 to easily know how much there is ; and both are much more accurate than previously.

You can look at my demo here :

The look may yet change before release, but the main improvements will stay.

My settings include notifications for "Your time has been beaten" ; "Your top 3 time has been beaten" ; "Your WR has been beaten" ; "New IL WR in followed game" ; "Your time has been beaten in a followed game" ; "Your top 3 time has been beaten in a followed game" ; "Your WR has been beaten in a followed game".

Despite this, I got zero notification when two of my IL WR where beaten in a followed game (STK).

Video treated ; now it only needs to be uploaded (that's taking time...) (EDIT : done !)

I raced without direct run comparison, only with my clock. It appears that I remained close to Kitoko until the 2nd GP. Here, I did some mistake and lost nearly 30 seconds to him.It grew slightly until Blackhill Mansion. I hate this shortcut, because it's quite random. Even after a lot of practice (and using the horse kart because it is the one with which I get the more consistent result here), I failed twice on three laps. Then, I did two good single race challenges ; and two excellent closing GP (I got blackhill mansion perfectly the second time). After that, it was a matter of not screwing up. The final time is at 58m12s. Splits : https://splits.io/2h56

@UbuntuJackson : I think you meant expert% ! And it would sure be cool to have a run from you in the 0.9.3 expert% leaderboard.

Personally, I find the GP skips interesting to do.

It bothered me to have a time I could yet improve a lot.

So I did some runs tonight.

I avoided having too much big mistakes, and managed an exceptional double-skip in the "at world end's" GP (the one with 4 races), thanks to which I can proudly announce that despite those freakin' loading times, I have a new 0.9.3 Expert% WR. With bunny ears on most karts...

I still need to do some treatment on the video. Then, I'll upload it for verification.

(Edited on 20/04/19)

I've stopped maintaining this changelog, but many elements are mentioned in the changelog here : https://github.com/supertuxkart/stk-code/blob/master/CHANGELOG.md

There is more details about most of the important changes in some of my blog posts over at http://blog.supertuxkart.net/

(Edited on 15/10/18)

The next version won't be 0.9.4 ; though I can't edit the thread title to reflect it.

It won't be 1.0 either, because some things (like a new overworld) won't be in the next release.

So it will be 0.10

0.10's release is currently blocked by the online multiplayer code. There are other stuff like updating tracks, standard replays, important bugfixes... which are not yet done, but they aren't what make the release take time. They wait because they are not the current focus.

I expect an official networking beta to begin by end of october/beginning of november ; and a release in early 2019 - while a december release would be possible, the project's lead will most likely not manage to do his part fast enough.

The big change of 0.10 will be a first version of networking, and the release date depends of this. This won't affect speedrunning itself, but it may be fun to play and attract new players, so that's good.

(The following list is NOT complete)

On gameplay/UI changes affecting single player :

• The new mansion track will replace the old (good riddance)
• The story mode bug when sometimes the points of a challenge were not counted if skipped too fast (which is very annoying when in a run) has been fixed
• Better menu UI scaling (there should be no more tiny icons in difficulty/race mode selection screen)
• Rescue has been improved : the kart won't appear out of nowhere on the road, there will be an animation before (useful to not collide with a rescued kart)
• The bubble shield will flash quickly when it's about to end
• For shortcuts fans, this will be more disappointing, as they are mostly patched out with new checklines. Some small are still here though (math's class shortcut between the chairs, Hacienda's sand shortcut, minigolf's shortcut)
• Ranking computation improved (kart won't be displayed first if it goes backwards after crossing the lapline)
• AI improvements in item/nitro handling (by me). The main AI weaknesses are in the way it chooses its path and skids, but this still make an important difference, as it previously used nitro in the worst way possible and had some glaring weaknesses in item usage (didn't know to use a shield/swatter to remove a bomb/parachute, for example).
• The new speedometer and nitro gauge presented in this thread have been integrated.
• The slipstreaming changes presented in this thread have been integrated
• Significant improvements to splitscreen multiplayer (possibility to have 5 or more players ; better scaling and placement of several UI elements).
• Fix handicap (several values fixes update, including diminishing the actual base speed and acceleration)
• The ghost mode improvements presented in this thread have been integrated. There could be further refinements (comparing more than two ghosts, the ability to switch from one ghost to the other), but it's already a huge step forward.
• The sound in nitro challenges has been fixed (will be triggered even if you used some nitro) and a sound has also been added for when you go from over the target to under it.
• Bigger minimap (+80%) for easier quick lookup, also displayed during the initial countdown phase.
• New visuals for skidding particles.

Other will certainly happen, I'll update this list from time to time.

The track's name in the IL leaderboard is incorrect (it should be Hacienda, not Hacendia) ; and Kitoko's time is off by 2 seconds according to the time in his video.

I put it in the comment of my level run but it seems to have be missed.

Okay. I stopped time on GNU appearing while you stop counting a bit earlier, before this last loading screen. I'll take note for the next run.

So does the in-game recorder work for this purpose ?

It's obvious that RTA without loading times is a better measure of skill than RTA with loading times ; so does Rph's solution work ?

@UbuntuJackson : why is the time for my latest run not the same in the leaderboard and in the video ? Have I missed something ? I thought putting the same times in the form.

Not that I'm against winning 2 seconds, but it's weird having different times displayed, and it's not like it affects the ranking anyways.

(By the way, the Hacienda's time of Kitoko is also wrong by 2 seconds)

Because triple skip means you have 0 points in the 3 races skipped ; and you can get a maximum of 12 in the two races you play by winning both.

EDIT : ninja'd

I finished a run five minutes ago. I beat my record despite an enormous mistake. I need to treat the video before uploading it.

Otherwise, no, I know nothing about Fouks, except what I've read on this forum. It seems he has deleted his account. That's sad for sure, I'd be curious to know too why, but we can't know.

"I think I won't do more full-game runs on this computer."

What bothered me was producing an horrible quality vid and having the run-time destroyed by loading time. But I get a decent quality if I record on Linux rather than on Windows. I'll try something.

EDIT : You can slash down loading time by preloading all the tracks in the RAM before the real run. Use a profile which has them all unlocked. Over the course of a full run, it can easily mean 30s saved if you have a computer like mine. (There still will be load time as the game need to render the track, but the "get it from the hard drive part" is at least skipped)

@Kitoko wrote in the GP double-skip guide that "It could be possible to triple-skip but that will probably require having psychic powers or an unfathomable amount of luck" because "the result of such a gran prix would have a 5-way tie for 12 points".

That's not correct.

Let's count the total points distributed : 6+4+2+1=13. 13x5 = 65. 5x12 = 60. 65 > 60.

However, double-skip is in theory possible in the 4-races GP, where the total amount of distributed points is 52. You get 12, so this let 10 per AI. Very small margin for error. EDIT : but definitely possible to do in practice