The idea for Coach came to me some years ago, when I was watching a pro football game on TV. I had an urge to play a computer football game on my PC, but it seemed somehow unsatisfying. What I really wanted was to build my own team, and coach it for years. The computer games around really only let you grab a joystick and run your little men around on the field, which was fun of course, but not what I was looking for at that time. Thus was planted the seed for these games.

So I sat down at my computer and started writing. Version 1.0 of Coach was almost entirely written in a single evening, with only a few bookkeeping routines that needed to be left for the following day. Compared with the amount of work that I have now put into Coach and its offspring games, that was pretty impressive.

Of course the original version of Coach was not anything like the current version. The first thing to be improved was the user interface, which forced you to go through your whole team every game and was prone to crashing when you entered an incorrect choice. The current look, with the blue screen and the full-screen menus, is probably the oldest part of the program that has survived reasonably intact.

Most of the major changes in the game have been on the GM side, only recently have I made major improvements to the coaching aspects. In Coach 1.0, the only way you could get a new player was to draft him. So the first large addition to Coach was to add trading. Making the computer a smart trader turned out to be one of the most challenging aspects of the game, far more so than, say, coaching a team. In fact I am still making modifications to it occasionally.

The next addition was primitive economics. Something to balance out the game a bit. The first economic rules had only player salaries, contract lengths, and a salary cap. Added at the same time was free agency (can you imagine Coach without free agents now?). Making the computer a smart bidder has probably been the #2 challenge in the game, and that is also being revised once in a while as well.

Very little changed in the game for a while, as I spent most of my effort putting together the other five games of the series: Manager, Basket, Hockey, College Coach, and College Basket. At first the development of all six games was basically independent, but when I ported everything over to C/C++ I started using shared code, so that a change in one game will affect all games that use the changed code. This has made the series more playable, as they basically all look the same. The last game to get converted to C was Hockey, which came out in 1996.

Most recently, there have been a large number of minor changes. The one major change has been finishing the economic rules. Instead of merely having a salary cap and a payroll, you now have a team-by-team budget based on the amount of anticipated revenue and savings before the season. The coaching options of all the games have been largely overhauled frequently, which was necessary with the economics. The problem was that the economics rules would tend to destabilize the game as the good teams could afford a higher payroll. So the redone coaching rules allow for a well-coached team to make up some of the difference.

The College Manager game, written in 1997, was the first new game in three years (the other college games came out in 1994), and was also the first of my games to be written more for e-mail play than for play on a home computer. Finally, the first new sport simulation since 1993 was the soccer game, completed in early 1999. The soccer game was a change of pace for a couple reasons. First, it was the first game to make its premier as an e-mail league (the home game didn't come out until after one full season had been played by the e-mail league). Second, it was the first game not based on American sports. The typical American division of the league in half (east and west in basketball and hockey, for example) was done away with in favor of a European system (all teams together, playing each other team twice). The cup matches also happen during the season, leaving the "real" championship (the team with the best record) to be possibly decided in the last game of the season. It has also been enjoyable, as the online sports leagues have attracted a much different group of people than the American football/baseball/basketball crowd. College Hockey was written in 2001, which like the other college games is largely an adaptation of the pro game.

I will start with talking about the general manager options, largely because they have been the most important part of the games throughout the development. The basic purpose of the GM functions in Coach (and the rest of the pro games of course) is to allow you to build your team in whatever ways are available to pro teams, and to make the game as playable as possible.

One tough case to balance is the free agent signings before the draft. Obviously the fairest way to run the whole thing would be to have every team, one at a time, choose the player it wanted to bid on and run a set of bids for that player. Then continue this until either no free agents were left, or more likely, no team wanted to make a bid.

Doing free agents in this manner would definitely be the fairest way of running it, but did not work out in practice, since the person in front of the computer would be wading through screen after screen of unwanted players to bid on before getting his chance. So out of that came the current method of free agency, where your team gets to choose players for bids first, and the computer teams sign right before the draft.

This aspect of the college game was basically nonexistent for a few years. All you did was decide what positions you wanted new players in, and your roster was randomly filled with players of those positions. So recruitment was added to spice things up a bit. But how precisely to do it was tough, and the main reason I waited so long to do it.

In the "real world", recruiting of high schoolers is quite a complex process. For starters, all players have some sort of innate preference for most of the schools. It has to do with where a school is, what conference they are in, where the kid grew up, where his folks went to school, how good the school's football team is, and how long before he'd start, among other considerations. There's really nothing that can be changed about that. So what the recruiters try to do is either overcome this preference or build on it. The colleges, for their part, can show varying degrees of interest in an athlete, from a mere letter to a visit to the school and getting the pitch from the head coach.

In putting this in the college games, I didn't feel that making it that complex would be very useful. The basic recruiting scheme of having a pool of athletes that the schools try to recruit was easy to come up with. The hard part was the mechanics. I finally decided to not give the players any preference up front, but instead to have them rank each team that was interested in them on a basis of the school's ranking and the quality of the conference. To make it fair for coaches, the conference only matters for computer teams. The remainder of a player's preference was put in the form of a random number added to each team rating, and the player would go to the school with the highest overall rating.

The aspect of "owner" of a team is quite new in the games. It used to be that the salary cap was the only thing preventing infinite payrolls. The game played OK that way, but was definitely lacking a dimension. So the full-scale economics rules were added, and later simplified for ease of play.

They are definitely simpler than the real-life ownership. There is only one ticket price for the whole stadium, for example, and no luxury suites. The TV revenue is a constant amount, no matter how good or bad your team is or how large or small a market you are in. In the way the rules ended up, the TV revenue is basically a force to equalize the teams, and the gate revenue is based on how good your team is.

Of primary importance, though, was forcing real-life considerations and common sense to rule. For starters, a winning team can afford a larger payroll. A team with a newer and larger stadium will bring in more money, but only if you fill it. A team that plays in an empty stadium loses much of its home field advantage. And if you start losing money, you need to cut your payroll.

A lot of fuss was made over the economic rules in the email leagues when I first introduced them. The claim was that I had forever ruined any hope of the worse teams rebuilding, as the better teams would have more money to throw around. Fortunately, that claim is completely false, as has been demonstrated by plenty of capable coaches in the e-mail leagues.

The rules on owners and general managers are basically the same for all sports, especially in these games where I have generalized the rules for my own sanity. So making each sport simulation really "feel" like the sport was quite a challenge. The obvious feature of all sports is that you decide who plays and who doesn't. The real question is what to do next, to give each game some character

In basketball there isn't really much I added. The problem is that doing detailed coaching, such as designing plays and such, would be beyond the scope of the game. Besides, plays in basketball are harder to work out than those in football, and it would have likely turned a simple, elegant game into a coaching nightmare for people who didn't want to spend too much time.

So I basically left it at choosing your starters, giving the players preferences when on the floor, and so on. While I'm on the subject of basketball, one of my favorite aspects is that there is very little difference between the positions in the program. Each is equally likely to get a rebound, take a shot if open, or make a pass. The only real difference is that the guards inbound the ball and that the center tends to be closest to the hoop on defense, then the forwards, then the guards. All other statistical differences come from who plays where and how you use them.

Why this is important is that a coach or GM has a lot of flexibility in building a team. Do you want to build around passing the ball to a big guy inside? Or do you want to have a small forward or shooting guard take the shots while your inside guys mainly do defense? Ideally, you would like every player at every position to be great shooters, passers, AND rebounders. But reality gets in the way, and one must choose what character team you want.

The game of football revolves around offensive plays to a large degree. The job of the offense is to get a play and formation sent from the coach and try to execute it as well as possible. Coach comes with many built-in plays, and you can design up to 55 additional plays. The plays are then put into five playbooks for different situations (normal, short yardage, long yardage, red zone, and goal line), and you can decide what plays are available in each and how often they are to be used.

The defense has a basic formation that it uses, but its real job is to try to handle whatever the offense sends at it. So I chose to leave out defensive strategies, and basically leave it at choosing starters, a formation, and how much man or zone coverage to use, and how often to blitz. You can customize your defensive playbook (as with offensive plays), which is usually used to create a larger number of defensive sets.

Baseball is probably about the simplest sport around. What makes designing a baseball game so hard is the number of little details that usually won't make a difference in one game (or in ten games for that matter), but will give you a couple more wins in 162. When writing the coaching routines for Manager, I again chose to take a middle road between simplicity and detail. So I left out choices like what pitch to throw when the count is 2-1 and the bases are empty. Instead, the basic choices are in there. How much do you want to bunt? Hit and run? Steal? Pinch hit? etc. In fact, Manager has far more preferences than any of the other games, so you really do get to decide how your team will play.

As far as the lineups, Manager is far more complex than either Coach or Basket, because you set your defense, batting order, and pinch hitters both for playing RH pitchers and LH pitchers.

One of the "fun" aspects of baseball, though, is that there's not really much you can do to help your team out. In football, you can switch between a rushing and passing attack or try a new formation. In basketball, you can change defense to double someone who's scoring too much. But it's hard to make a certain batter or pitcher perform better. So this is why I have put so many preferences and player options that you can tweak to get the best performance out of your team.

Next is hockey. On the surface, hockey was basically a port of much of the basketball code with major revisions. The game is, like basketball, essentially passing until you can get a good shot or lose the puck. But below the surface, there are major differences. The critical aspect to coaching hockey is managing your lines well. This is certainly an art and not a science, similar to putting the perfect batting order together in baseball. There are many jobs that need to be done on the ice, and at least one player must be capable of each of them. The obvious things are shooting, passing, checking, rebounds, and faceoffs. Like in basketball, in a perfect world every player could do everything. But like in basketball, this isn't a perfect world...

Finally is soccer, the "baby" of the family. Like basketball, the soccer game turned out to be a very flexible game, where you can assign your players various tasks on offense. The trick here is to come up with a game plan that will make the best use of your players. You can set how hard you want to press attacks, how much defense you want to play, your formations, and numerous other options.

If you have ever attempted to write artificial intelligence routines for something other than chess or checkers, you probably quickly came to the realization that it's quite difficult. I think I have done a pretty good job with the computer coaching and GMing, and it doesn't make stupid trades with anything near the frequency of, say, FPS baseball '96. Even so, a good player can routinely beat up on the computer. And beside, wouldn't it be more fun to be playing against a real person? So a few years back I proposed starting an email league based on the Coach game. That league has been quite successful, and in fact there are leagues based on all nine of my games. And yes, it is much more fun beating up on teams belonging to real people than it is playing the computer...

Really, though, league play is quite involved, and a much different character than sitting in front of your PC and playing a season of Coach. League newsletters have come up, the college leagues have coaches' polls, and you get to deal with real people when making trades.

One side effect of the leagues is that the ability of the computer teams has really been stretched to its limit. The quality of a person's coaching and GMing is much higher when there are only two games per week, and all the stats are available on the WWW.

Where to from here? This is always the question I ask myself after a major update of some type. And numerous times I have decided that the games had arrived to their final state, only to add something new because of a comment from someone or a brainstorm on my part. So the final answer may be that the games will never be "finished". But I do believe that the period of frequent major change is over, and that future upgrades, while not insignificant, will not be major reworkings either.

One thing that has kept the games progressing to a better and better product is feedback from users, who often have different ideas of what they would like to see in a game, and often those are things I have never thought about. I do hope that will continue in the future, as this feedback is responsible for nearly half of the major improvements in the games so far.

Andy Dolphin