The Economy of Blood Bowl

[Deathwing]

Had chance to have a look. Wow. Impressive bit of kit. Thanks H&M!

[GalakStarscraper]

The trouble of on-line communication. I know i've inadvertently rubbed a few people up the wrong way both here and on the GW forums.

Maybe we should have a sarcasm emoticon here as well as the rest.

I also used
[sacrasm on]Beware the HALFLINGS, Darkson, muh HA HA HA[sacrasm off]

Works for me, and I've used it before

Galak

[High & Mighty]

The trouble of on-line communication. I know i've inadvertently rubbed a few people up the wrong way both here and on the GW forums.

Maybe we should have a sarcasm emoticon here as well as the rest.

Sometimes you just have to remember it's a game of little men before taking things too personally. I'm glad some people liked the sheet. No glaring mistakes then.

[Darkson]

The trouble of on-line communication. I know i've inadvertently rubbed a few people up the wrong way both here and on the GW forums.

Maybe we should have a sarcasm emoticon here as well as the rest.

I also used
[sacrasm on]Beware the HALFLINGS, Darkson, muh HA HA HA[sacrasm off]

Works for me, and I've used it before

Galak

yeah, yeah yeah, rub it in why don'y you.

[dreaming mode on] I'll get you next time.[/dreaming mode]

[Mestari]

With team management becoming a more and more important part of Bloodbowl, this begs the question what exactly does a coach need to do to develop the team they want.

I finally got time to review this thing.
The spreadsheet seems like a very detailed and well-made tool.

However, this is hardly a handy tool for most of the coaches. The problem being the amount of calculations bestowed upon the coach when the spreadsheet could've made it instead. Another problem is the team management decisions that are now controlled by a set of rules.

This is what I'd be after (listing of the worksheets):

Assumptions and info-sheet (self-explanatory)

Input-sheet:
This sheet contains a table for inputting the starting variables (player costs, amount of players etc.)
Also, below that table, it includes a table for the following purpose:
Instead of calculating any percentages, the coach inputs every game that the team has played into the spreadsheet, giving all the necessary information for the spreadsheet to calculate win percentages etc. The list would include the amount, cost and spp's of retired players, amount of money earned, ff change etc as the columns, and each row represents a single game.

Data-sheet:
The spreadsheet then sums up the games into a table similar to the one that we are now forced to fill in ourselves. It will also include additional things like the average SPP's of a retired or a dead player etc.

Diagram-sheet:
This sheet includes the diagrams found in the current spreadsheet plus some additional things, as in addition to offering an easier and less error-prone way of using the spreadsheet, the above way of inputting data would allow certain operations that cannot be reached by the presented spreadsheet:
Namely, it doesn't lose some details that can be used to make additional observations. The fact that we'd have input at an accuracy of one game, we could make calculations about second derivatives (And third etc ) - we could maybe even give up the use of averages in the calculations and use formuli instead, where the variable used is the amount of games played.
Naturally I understand that making calculations with second derivatives is rather inaccurate as long as there is only a small number of games, but using assumpted averages is hardly more accurate (unless we have a reliable source for getting the averages).

What do you think, H&M? Your spreadsheet is a better one to study the long-term trends as you can simply decide the averages etc., but the one suggested above would be more accurate what comes to a single team and help in assessing the development of your team.
If there'd be an option of disregarding the n last games, and thus to view the development of data and the figures, this could be extremely useful...

If I'd know how to dynamically do graphs to php homepages, I'd suggest we'd do it as a publicly available web-site, but alas, I don't know.

[High & Mighty]

Hey. Thanks. Below are the responses to each of your points.

However, this is hardly a handy tool for most of the coaches. The problem being the amount of calculations bestowed upon the coach when the spreadsheet could've made it instead. Another problem is the team management decisions that are now controlled by a set of rules.

The point of the spreadsheet, though, is to model the team going forward. Not just calculate what they did in the past. It did start out much simpler and just tended to long-term trends after hundreds of games, but wasn't really all that interesting in any real world setting which is why I added the management aspects.

Regarding the management decisions governed by rules, that was the cost of avoiding what I thought was much more effort doing what I think you propose below.

Input-sheet:
This sheet contains a table for inputting the starting variables (player costs, amount of players etc.)
Also, below that table, it includes a table for the following purpose:
Instead of calculating any percentages, the coach inputs every game that the team has played into the spreadsheet, giving all the necessary information for the spreadsheet to calculate win percentages etc. The list would include the amount, cost and spp's of retired players, amount of money earned, ff change etc as the columns, and each row represents a single game.

I may be missing the point, but it would seem to be much more work entering the results of every game you have played and any transactions you have undertaken rather than just saying you win 60 percent of your games and a player retires every 7 games. And again this would only handle the past events. You would still need to say something about the future. Otherwise you are just getting graphs of what you have already done.

Data-sheet:
The spreadsheet then sums up the games into a table similar to the one that we are now forced to fill in ourselves. It will also include additional things like the average SPP's of a retired or a dead player etc.

I did protect the spreadsheet, but I didn't lock it, so if you go to Format-->Sheet-->Unhide, and select the sheet 'FF' you will see where the simulation is taking place (wasn't really meant for human eyes so largely a lot of numbers). Every game result is being reported, so essentially I'm doing exactly what you want but in the opposite direction.

My thinking is that it is interesting to see where your decisions could take your team down the road. Certain things, like what percentage of your games you will win/score 2TDs/2 Cas, have an impact in limiting the maximum development of your fan factor. Other decisions, like how frequently to retire a player, determine where your TR is capping out at. Yes in real life a coach will decide between retiring a player in game 17 or 18 based upon whether he could get an extra 10k from it or another handicap roll, but the point of the current rules is that you will have to retire players on a continuous basis at some point in order to avoid zero revenue.

So the question becomes, ok, starting out with a TR 100 team, I will max out my roster after 12 games, get to my desired rerolls after 17, so now, how frequently do I have to retire players in order to keep my average revenue at 40k per game. So you can play with your rate of retirement to see what is happening to your TR. If you aren't retiring enough, it will keep going up. Too much retirement and it will plummet. Just right, and you get a constant TR.

The fact that we'd have input at an accuracy of one game, we could make calculations about second derivatives (And third etc ) - we could maybe even give up the use of averages in the calculations and use formuli instead, where the variable used is the amount of games played.
Naturally I understand that making calculations with second derivatives is rather inaccurate as long as there is only a small number of games, but using assumpted averages is hardly more accurate (unless we have a reliable source for getting the averages).

As I pointed out above, the sheet does calculate everything on a per game basis, but since it's a simulator, it does this with the person entering what they expect/desire/hope will happen. The formulas that determine team development are fairly straightforward and require only the information asked for on the sheet. (Granted, some of these such as win percentage are determined by lots of other factors, but that is why I ask for a number rather than pretending to create an equation that calculates it). There are formuli underlying everything, and (largely because of a mathematical problem below) done on a per game basis, but I didn't think anyone would want to see these.

One mathematical problem is that the negative fan factor modifier for 10+ FF is a discontinuous function (if you have a FF=15, you have a -1 modifier, not -1.5) so while it can be represented in an equation, the equation cannot be simplified beyond that form. This is easy to do with IF statements in excel, but it creates a roadblock to simplying the underlying equations on paper. I did think about putting some results on the sheet regarding the rate of convergence and growth of certain variables. I left it out for now largely because I needed to do some real work for a while, but derivatives are not out of the question.

What do you think, H&M? Your spreadsheet is a better one to study the long-term trends as you can simply decide the averages etc., but the one suggested above would be more accurate what comes to a single team and help in assessing the development of your team.
If there'd be an option of disregarding the n last games, and thus to view the development of data and the figures, this could be extremely useful...

If I'd know how to dynamically do graphs to php homepages, I'd suggest we'd do it as a publicly available web-site, but alas, I don't know.

When I was making the sheet, the goal was something that you could enter basic data and see how the team would evolve over time. If you are going to play a team with a higher player cost, bump up the roster cost by 200k and see what happens if everything else remains the same. Change FF from 1 to 9 and see what happens. Have 4 players die in game 14 and see what happens. Don't win any games from number 4 to number 20.

If you want to put in an existing team, rather than a new one, just enter in your current team and the expectations for the future and it spits it out. What I don't see is how entering actual data for the past makes the projections for the future development any more accurate or easy. I think I need that part explained further.

Now that I've said that, when I made it, I did think about whether it would be possible to have the spreadsheet pull in data from an existing source. For example, if you use Dangerous Dave's team sheet, it could pull the data from there and put the actuals in. At that point, you could have the graphs show your team's development if everything continued according to the past, and then, based upon your predictions for the future, what's the alternative (so you'd get wishbone lines). I just couldn't see if there was any value-added there. Plus I could see that this wasn't less work, and in fact, is likely more work. And most importantly, real work got real busy...

So my basic question is, what is the value of actual historical data for projecting the future. Or is what you want just something that reports graphs/derivatives/etc on how the team has developed in the past?

[Mestari]

What I don't see is how entering actual data for the past makes the projections for the future development any more accurate or easy. I think I need that part explained further.

Simple example:
Assume that the column TDf includes how many touchdowns the team has made in each of their games.
By taking the average of the values in the column you get the average TD's per game, that would serve as the input data to your table instead of guessing or predicting values for this. By counting the percentage of the games with 2+ TD's you get that input etc.
Also, it does allow treatment of the data to examine the success of the team so far, and how, according to this data, the team will probably keep playing in the future.

So my basic question is, what is the value of actual historical data for projecting the future. Or is what you want just something that reports graphs/derivatives/etc on how the team has developed in the past?

I'm not sure if I explain myself correctly, but the point would be to use the past success of the team to predict automatically the values that are now to be guessed or predicted. In addition, as the management decisions can be quantified from the past decisions of the coach, there is no need for rules to predict future decisions.

This vision serves the purpose of managing your team. Your spreadsheet is definitely better for examining what would happen _if_ something was done differently, games were won by a different margin etc. I'll see if I manage to get some extra time to make an example.

[High & Mighty]

Simple example:
Assume that the column TDf includes how many touchdowns the team has made in each of their games.
By taking the average of the values in the column you get the average TD's per game, that would serve as the input data to your table instead of guessing or predicting values for this. By counting the percentage of the games with 2+ TD's you get that input etc.
Also, it does allow treatment of the data to examine the success of the team so far, and how, according to this data, the team will probably keep playing in the future.

OK. But now in order to do this, you have to already have the information on every game you have played. In this case, either you've got to enter all this information again in this spreadsheet, which takes alot more time than entering one number, or, since you already have all the data somewhere else, you can very easily take the average from that data there.

The ability to show what your past experience has been would be nice, but this would likely only be possible if I were to have the workbook look at another file and get the information from there. Entering it all over again would seem like a very large waste of time when you can enter just a couple numbers to get the same information. But that way it could graph both your current trend and an alternative based upon the values you decide to change.

I'm not sure if I explain myself correctly, but the point would be to use the past success of the team to predict automatically the values that are now to be guessed or predicted. In addition, as the management decisions can be quantified from the past decisions of the coach, there is no need for rules to predict future decisions.

But using the past experiences of the team doesn't accurately predict the experiences in the future. Yes, you can show a graph showing "If you continue at the current rate, then..." but players may die less in the future and you will have to retire players faster in the future than in the early development of a team. You will still have to say the max number of roster spots and rerolls you want to have and which you want to buy when. You either will or will want to to earn SPPs at a different rate.

And as you stated before, trying to take the derivatives of just 5 or 6 games are going to be of questionable value. And so you have to have played many more games in order to get reasonable approximations, which means more work in entering game by game data. So you can extrapolate future development from past experiences, but I'm skeptical as to whether this would be either more valuable or more accurate.

[Mestari]

Matter of preference, I guess. I feel that both of them would be useful and having them both would increment the amount of value that can be extracted from them.

Also, should you be able to gather large amount of information, having the data entered in chunks of one game at a time, you could use it to make more accurate predictions about the future trends. For example the concepts of "early spp development" and "spp development after that" seem rather unfounded. Where is the data that shows that this division can actually be made? By examining large amounts of data we could confirm trends such as these. All in all, to study the statistical phenomena in this game, I believe that having a sheet something like I suggested would be more useful. Which is not to say that your sheet isn't useful - but it only helps in predicting where your team would end up given these assumptions about your future success, whereas the game-at-a-time data would allow studying what these predictions should be. In addition it would help to define where your team will end up if the variables counted from the previous game develop with the same rate that they have developed so far using least-square fits etc. Naturally I realise that such fits would not be very accurate in the first few games, but I'm not concerned about teams that only play a few games.

[High & Mighty]

By examining large amounts of data we could confirm trends such as these. All in all, to study the statistical phenomena in this game, I believe that having a sheet something like I suggested would be more useful.

This is probably something more in line with the other project I was working on before this stole my attention for a little bit. Galak was nice enough to add an option to the PBEM software for outputting a log file and so I'm trying to put together something that will analyze your PBEM game. That way, you've got per game analysis of the events in each game so over time, you'll have a sizeable repository of data. Unfortunately the evil work people keep demanding real work get done.

For example the concepts of "early spp development" and "spp development after that" seem rather unfounded. Where is the data that shows that this division can actually be made?

This is something that comes in in team development. If you don't think you will score less or more SPPs later, make them the same. But personally, as my own team developed, it would reach a point where I would try to reduce overall SPPs and spend more time trying to target them to the few underskilled players I have. It's just a strategy I would use since beyond additional skills, more than 2 TDs or 2 Casualties or any completions for that matter are wasted SPPs that only serve to inflate your TR and reduce your winnings.

[Mestari]

This is probably something more in line with the other project I was working on before this stole my attention for a little bit. Galak was nice enough to add an option to the PBEM software for outputting a log file and so I'm trying to put together something that will analyze your PBEM game. That way, you've got per game analysis of the events in each game so over time, you'll have a sizeable repository of data.

I'm looking forward to seeing that.

This is something that comes in in team development. If you don't think you will score less or more SPPs later, make them the same. But personally, as my own team developed, it would reach a point where I would try to reduce overall SPPs and spend more time trying to target them to the few underskilled players I have. It's just a strategy I would use since beyond additional skills, more than 2 TDs or 2 Casualties or any completions for that matter are wasted SPPs that only serve to inflate your TR and reduce your winnings.

Ok, I misunderstood that one.