"This is a zoomed-in view that constrains the graph domain to only the attribute values that the MvM sentry firing speed upgrade is concerned with. To zoom out, go here.

The dotted lines indicate the "ideal" firing interval for particular values of the "mult_postfiredelay" attribute. The solid lines are stairstepped, showing that in reality, the game has to round the firing interval to a discrete integer number of game simulation ticks, which can lead to counterintuitive results.

Sentry firing speed is worse than minigun firing speed because not only are the intervals rounded to integer numbers of ticks, but they're actually rounded to a multiple of three ticks, since sentries are only simulated every third game tick (22 times per second). This makes the stairstepped "actual" values very different from the diagonal "ideal" values in many cases. It also means that there is effectively a hard cap on the max sentry firing speed: it cannot ever fire faster than once every 3 game ticks (22 times per second), even if you have max firing speed, are using the wrangler, and use a crit canteen.

Notice that the jump from 9->6 for unwrangled sentries and 6->3 for wrangled sentries both happen almost exactly at 0.9000 (which corresponds to +1 firing speed upgrade). Because the transition happens extremely close to 0.9000, you will usually get the quicker firing interval with just one firing speed upgrade. But if the server's simulation tick timing is slightly off, even if a tick happens just 1% later than it should (15.15 milliseconds instead of 15.00 milliseconds for example), then you may end up with the slower firing interval. This is why it is sometimes but not always necessary to buy the second firing speed upgrade to actually make the sentry fire faster. It's very inconsistent and temperamental.

On the x-axis:

• 1.00 = no firing speed upgrades
  
• 0.90 = +1 firing speed upgrade
  
• 0.80 = +2 firing speed upgrades
  
• 0.70 = +3 firing speed upgrades

"