BuzzerBeater Forums

how is the salary efficiency calculated? I have a SF with nearly 0,7 whereas my PF hast 0,3 even if i trained him a lot in secondaries.

I am just curious how it works and what kind of efficiency is good for each position

It is an experimental metric. What it does is divides actual salary by an "average" player salary calculated by averaging the coefficients for every skill and position and using TSP as the determining factor. My experience is that it isn't terribly useful in comparing players of vastly different salary or position. But it helps distinguish players with the same position and/or similar salary. I would be curious whether or not you find it useful. I use it to determine what is a "salary efficient pop" on a player I am training, but not much else.

Hi Rhyminsimon, and thank you for spending your spare time creating such great tool and sharing it with others. I've just started using your tool and it looks very nice. A question from me: what can be the reason that the difference between a full season of Passing (PG) and one of Passing (PG/SG) is just very minor (0.7 more or less)?

I simulate the following:
3 trainees, all 18 currently but I'm adjusting for the next season where they will be 19. Trainee 1 and 2 are 6'0" and Trainee 3 is 6'5"

PA Skills at start of season (assumed, not actual):
Trainee 1: 7.5
Trainee 2: 2.5
Trainee 3: 7.5

PA Skills after 1 season of PA (PG/SG):
Trainee 1: 13.42
Trainee 2: 9.46
Trainee 3: 13.29

PA Skills after 1 season of PA (PG) instead of (PG/SG):
Trainee 1: 14.18
Trainee 2: 10.37
Trainee 3: 14.06

Is this an expected outcome? I thought from reading the training speed analysis topic that single position PA training would amount to 1 pop per 1-2 weeks, while double position training would amount to 1 pop per 2-3 weeks. Is this difference just that small (1.9 weeks/pop vs 2.1 weeks/pop), is the simulator not accurate or am I making a mistake in using the simulator?

Thanks for your time

Thanks for bringing this to my attention. When I first created this three months ago, I was working with the best numbers/analysis that was available at the time. Since then I've concentrated on adding functionality and extra features, and assuming that the numbers/analysis hadn't changed my model much.

At that time there was some noise in our data which skewed the computer's analysis of Passing (PG). It suggested an unrealistically steep slope (like .3 for a 18yo 6'0" & 1.1 for a 18yo 7'6") for Passing (PG). In order to better approximate reality, I talked with wozzvt (who runs the BB-Training site:(http://training.bb-usa.net)) and we estimated the Passing (PG) numbers by looking at the all heights <6'8" (.60) and >6'8" (.63) which I used in my model. (In the interest of full disclosure, this and Shot Blocking (C) were the only training types I did this kind of thing with.)

Now that we've collected more data, the noise has quieted down some on this one. And when I look at the analysis now it looks like a 6'0" 18yo would get .73 from Passing (PG) and .53 from Passing (PG/SG). The slope still seems a little steep, but more reasonable. And the difference from the current model is not huge (at least for guards). But the analysis is closer to the order you were expecting than the current model.

Before you brought this up I had hoped to put off a new release until Season 23. But it looks like I should do it sooner. I will look at the raw data and the analysis and then: I will update the model this week or next and release a new version.

In the meantime, here's a quick fix using the computer analysis displayed on the BB-Training Site currently:
– Go to the "Coef" sheet and in the top left you will find a table labeled 6'0",18yo.
– In the PASSING (PG) row find the PA column and enter .73.
– Directly below find the PASSING (PG/SG) row and enter .53 in the PA column.
– Find the table labeled 7'6", 18yo below.
– In the PASSING (PG) row find the PA column and enter .88.
– Directly below find the PASSING (PG/SG) row and enter .64 in the PA column.

Last edited by rhyminsimon at 1/24/2013 9:37:55 AM

Thank you for your quick and thorough response. I was just wondering what caused the difference in my expectation and what the simulator showed, it could very well have been that I didn't know the program well enough yet. If it wasn't, I thought I could just help you with improving the simulator.

Thank you for the solution offered, and don't feel obliged to change your plans and update sooner, I'm already glad to be able to use your program for free!

I've considered this. Before adding [EDIT: any representation of the cross-training effect] to the model, I want to be sure that it is accurate. It may be that there is more going on with the cross-training effect than what you propose. From what the BB's have said, this seems to be the place where penalties are assessed for unbalanced players and possibly where benefits are awarded for balanced players. It's worth figuring out this before adding it to the model.

But rest assured, smart people are working on it. In the meantime, you are surely welcome to go into the tables on the coef sheet and add .01 everywhere if you want to adjust the model on your TS. Though don't blame me if your player models are a little inflated. ;-)

Last edited by rhyminsimon at 1/29/2013 4:05:48 PM

Actually I don´t think adding 0.01 would do any good.

If I understand the Crosstraining Concept correctly (which I doubt, as I´m not sure anybody does), it redirects 10% of the training towards other categories. But that means what?

If you trained Passing one position for a small player, you might get 0.45 training effect for a full week. Which is...how much? 100%? Unlikely, as there´s already some B-Skills down. So let´s say that´s 90%, so 1/9 from that (which equals 0.05 "Passing") will be shifted somewhere. And now all those questions ... is 0.05 apples the same as 0.05 oranges? Is all that elastic and size and age again considered, when applying it on another skill, making 0.05 apples all of a sudden 0.10 bananas or 0.0085 melons? How about 2, 3 positions training, and team, and...

I choose to sit and wait for the numerical analysis

Absolutely. That's my point. I considered doing something similar to this. But I'm not going to change the model for the simulator until there is a thorough analysis of the situation. The Training Simulator's reliance on the BB-Training Site's verifiable data and analysis is what makes it useful to folks.

That said... If someone wants to adjust the numbers on the coef sheet, they're welcome to. But I wouldn't recommend it. Nor would I do it myself.

One of the projects that the BB-Training Site is looking into is the cross-training effect. And when there's reliable numerical analysis, it will be presented and I will include it in the Training Simulator.

Last edited by rhyminsimon at 1/29/2013 4:04:32 PM

The aforementioned updates are now available! The TS is now using the latest data from (http://training.bb-usa.net/). Enjoy.

Thanks a lot for the update, I do have some more questions though:

Reading to coefficients, big players train passing faster than small players; is that correct? It seems a bit odd to me.
If you have a tall player, training 1v1 gets you JS pops faster than training JS?
If you have a tall player and you train HA, DR pops faster if you do 2 position instead of 1?