My goal with this guide is to figure out whether using arms-up while on the ground has an effect on your speed. To test this, I ran a number of test tracks both with and without arms-up pressed.
I ran two series of tests, one with different track inclinations and with / without fans or boosters, and a second series focussing on the effects of different speeds.

I did NOT test the impact of the following parameters:

• curves
  
• surfaces other than the default surface
  
• changes in inclination
  
• impact on anything other than speed (like e.g. handling)

First, lets state some base parameters that remained true for all tests:

• All tests were done over a straight track with a start- and end tile, with 14 road tiles between the two
  
• For the up- and downhill tests, the entire track was angled at +/- 10°
  
• For the normal test, no inputs were made for the entire test
  
• For the arms-up test, arms-up was pressed before the start and let go after the finish. No other inputs were made

These were the track layouts used for the tests:

My first test series was designed to test the impact of fans & boosters as well as slopes. These were the results from that series:

First of all, any impact made by using arms up is tiny: The biggest impact in this series is at -0.38%. Over a hypthetical 45 second race, this would come out to a difference of 0.171 sec. Sure, its there, but it shoudn't be your first concern on a new track.

The impact of using arms-up also isn't as consistent as I would have expected. In some cases, it is even faster to use arms-up than not using it. I am currently unsure why that can be the case.

Generally, the speed of the cart seems to have a sizeable impact on how much arms-up affects the cart. To go into more detail on this, I ran a second series of tests, focussing on the speed of the cart.

I only tested this for the simplest case, a flat track without boosters or fans, so I wouldn't have to run dozens of tests.
To achieve this, I placed a slowster immediatly after the start block, and adjusted the speed and strength setting of it to achieve the desired speed.
These are the results from that second series:

As becomes pretty obvious in the chart, the faster the cart is moving, the smaller the impact of using arms-up becomes. This might also explain the results of the first series, where there was little impact for the up- and downhill tracks. These tests necessarily had a relatively high speed, either to get the cart up the hill, or due to it accelerating downhill.

The bumpiness of the graph in the 100 - 150 speed range can be explained by measurement inaccuracies and rounding errors, since the race times in-game are only displayed to 0.001 sec.

During my testing, I came upon a few configurations where using arms-up was actually FASTER than not using it, usually by 0.4% - 0.6%, which is huge compared to the usual losses using arms up. I observed this for multiple configurations, even if very similar configurations behaved as expected. The effect was also consistent for each run if it appeared.

One example is the following track:

• start block: booster force 8, booster speed 150
  
• slowster, force 10, speed 120
  
• road block, 13x
  
• finish block

This setup resulted in a speed of 116 and the following times: without arms-up, the track takes 7.633 seconds, while if using arms-up, the track only takes 7.601 seconds. This means using arms-up is about 0.42% faster than not using it.

This is especially notable, since very similar configurations result in the expected 0.1% decrease in speed, rather than an increase.I have no clue what the deal is here, though it appears to have something to do with the slowster. When I release arms-up after the booster, the resulting time is the same as if using arms-up the whole way.

If anyone knows what is going on here, please let me know!

Generally, using arms-up on the ground will lead to a slight time-loss in most situations. The effect is more pronounced at lower speeds, and basically disappears above speeds of ca. 100.

However, there are some wierd mechanisms which can actually make it faster to use arms-up than not, though I do not understand them yet.

In total, I would generally recommend not using arms-up on the ground, though it shouldnt make a difference either way unless you are really optimizing a track. If you are trying for wr on a track, it might be worth testing if any wierd speed increases due to arms-up are present, especially around slowsters.

If you have any questions or comments, please put them below. Also let me know about any mistakes, I'm sure some snuck through.