Because conventional shark wisdom dictates that you wait on quarterbacks, most of you reading this will probably not end up with Peyton Manning on your team. Why spend your first or second round pick on Manning when you can wait several rounds and get production from one of the various Jakes or Drews that --- while not quite as good --- is at least in the same ballpark? Meanwhile you got a great running back or receiver with that early pick that you didn't use on Manning.

I am generally in agreement with that philosophy, but there is something to be said for the set-it-and-forget-it production you get from Manning or, to a lesser extent (at a cheaper price), guys like Tom Brady and Matt Hasselbeck. If you have Manning, you don't really need to worry about your backup too much, which essentially frees up a roster spot. And, usually, you do get more production out of the quarterback position than your opponents do. It may or may not be enough extra production to warrant the high pick used, but it is very likely that the Manning owner this year will have an advantage at the position.

As I think my way through problems like this, I like to generalize. Forget Manning. Forget Brady and Plummer and Delhomme and Bledsoe and Brees. The theoretical problems here are, for example, these: would you rather have QB1 and QB24, or QB12 and QB13? What about QB6 and QB20? How much do you lose if you go with QB14 and QB20? What if you like to acquire Jakes and Drews in bulk: would QB8, QB14, and QB17 give you a better chance at elite quarterback production than QB4 and QB15? Do you have any prayer if you try something really crazy like QB18, QB19, and QB20?

To investigate questions like these, I've built a model. Like all models, it's imperfect. I think it's got some value anyway, and I hope you will too. Let me describe it for you by working through an example. Let's say you've drafted QB6 and QB19, for instance. As I type this, that would be Donovan McNabb and Philip Rivers, but that's not really important. If you draft a "typical" QB6 and QB19, how many points can you expect from your quarterback position this year? That's the question. I'll make the following assumptions:

1. For the first three weeks of the season, you start the higher-drafted player: QB6 in this case.
2. Thereafter, you start the guy with the highest year-to-date fantasy point total.
3. You don't start anyone who is on bye or injured (obviously).
4. If both your quarterbacks happen to be injured or on bye, you get 10 points for that week. I am assuming that in most leagues in most weeks you can pick someone up who will get you ten points.

As I said, the model is imperfect. It doesn't take into account matchups, hot and cold streaks, or other factors that might affect your who-to-start decision. Those things could be built into the model --- and maybe someday they will be --- but for now I'm trying to keep it simple. You start the guy you drafted first (if he's not on bye) until he gets hurt or until it becomes clear that he's not as good as the guy you drafted later.

Now, we go back and look at how many points you would have gotten from your quarterback position if you had drafted QB6 and QB19 in previous years. Last year, QB6 and QB19 were Kerry Collins and Steve McNair. Here's how it would have worked out:

2005 week  1, start: Kerry Collins         25.4 points
2005 week  2, start: Kerry Collins         18.2 points
2005 week  3, start: Kerry Collins         25.1 points
2005 week  4, start: Kerry Collins         10.8 points
2005 week  5, start: Kerry Collins         25.1 points
2005 week  6, start: Steve McNair          11.6 points
2005 week  7, start: Steve McNair          17.8 points
2005 week  8, start: Kerry Collins         22.8 points
2005 week  9, start: Kerry Collins         15.8 points
2005 week 10, start: Kerry Collins         20.5 points
2005 week 11, start: Kerry Collins         17.2 points
2005 week 12, start: Kerry Collins         17.6 points
2005 week 13, start: Kerry Collins         14.8 points
2005 week 14, start: Kerry Collins         12.1 points
2005 week 15, start: Kerry Collins          9.7 points
2005 week 16, start: Steve McNair           0.5 points
2005 week 17, start: Kerry Collins         28.6 points
-------------------------------------------------------
TOTAL                                     293.6 points

This duo was a little disappointing, but fairly typical. From 2000--2005, the QB6/QB19 combination would have gotten you an average of 306.6 points. So how does that compare to the production obtained by someone who takes a stud quarterback? Here are the average point values for a few higher-priced QB options:

QB1 / QB8      340
QB1 / QB15     316
QB1 / QB24     315
QB2 / QB10     337
QB2 / QB19     337

Now it just so happens that QB2 has outperformed QB1 more often than not during the past six seasons, so the numbers seem a little strange. For example, the chart shows that a QB2/QB19 combination has historically done better than a QB1/QB15 combination. This should not be interpreted to mean that you should actually expect better production out of a later-drafted combination. It's likely a statistical fluke, and I would expect it to correct itself with another several seasons worth of data. But it does provide some confirmation of what we already suspected: often, you can wait to draft a quarterback still get one that's just as good as (maybe better than) the one you passed on.

Are you one of those people who likes to be the last to draft a first quarterback but the first to draft a backup? Such a strategy ought to land you with a QB12/QB13 combination, which has been good for an average of 298 points historically. Do you like to wait forever? A QB18/QB19 combination has been worth an average of 273 points.

Here is the point: no matter how good your quarterbacks are, it is unreasonable to expect more than about 350 points out of the position during the season. If you wait until a handful of quarterbacks have been selected before taking yours (e.g. QB6/QB19), you might expect to end up with around 306 points. If you wait a little longer (QB12/QB13), it's reasonable to expect just under 300. If you wait a really long time, 273 is a reasonable expectation. The difference between 350 and 273 is nothing to sneeze at. If you wait, you're in a bit of a hole. But the theory is that you can easily make up that ground (and much more) on other positions by not using a high pick on a quarterback. I will discuss that further at the end of the article. The main point here is that we have quantified --- at least roughly --- the size of the hole.

Because I love cramming a lot of data into a small space, I have created this chart, which gives the historical average points generated by each quarterback pair. Read your first quarterback along the top, and your second quarterback down the left. Go to the intersection, and you'll find the expected point value of that pair.

     1   2   3   4   5   6   7   8   9  10  11  12  13  14  15  16  17  18  19  20  21  22  23
 1
 2  347
 3  347 336
 4  352 348 333
 5  331 346 340 305
 6  345 349 345 323 322
 7  340 347 321 303 308 322
 8  340 373 345 326 320 328 310
 9  329 359 330 300 280 313 307 326
10  342 337 329 317 314 334 305 335 314
11  330 346 327 289 273 330 307 331 282 307
12  327 343 324 310 285 312 304 330 298 305 295
13  353 354 338 311 307 308 313 327 298 312 304 298
14  347 334 327 312 309 331 310 338 300 320 314 283 309
15  316 332 319 288 274 325 297 314 289 302 274 269 281 287
16  334 339 320 292 278 321 306 329 263 308 280 274 286 292 254
17  325 343 320 295 285 312 295 329 260 307 302 281 293 289 273 245
18  330 347 336 304 286 307 319 332 280 321 291 293 302 308 285 259 266
19  314 337 331 303 273 306 294 324 278 310 296 269 290 290 268 259 270 273
20  321 336 320 286 266 310 300 326 274 308 278 255 296 288 254 237 262 259 263
21  327 338 324 298 277 311 301 326 268 307 275 266 292 298 257 243 260 259 265 223
22  313 337 314 288 265 300 299 328 257 305 274 255 283 285 263 232 250 238 258 224 245
23  328 330 316 289 276 301 294 331 276 308 285 279 279 302 264 237 258 247 274 245 247 222
24  315 335 325 286 256 302 306 330 260 302 272 269 270 290 263 236 254 257 266 234 249 220 226

Now, what about that third quarterback? How much does he change things? Can you make up for a lack of quality by opting for quantity instead? Here are the point values for a few three-QB platoons:

QB6/QB19/QB20      310
QB12/QB13/QB16     300
QB18/QB19/QB20     275

As you can see by comparing to the numbers for the QB6/QB19, QB12/QB13, and QB18/QB19 pairs above, adding a third quarterback means almost nothing. So while quarterback bargains seem plentiful in the QB6 -- QB14 range, apparently you should not count on a quarterback drafted after QB15 to offer better production than not just one but two higher-drafted QBs.

Caveats

• Just for the record, I ought to mention explicitly that this article applies only to start-one-QB leagues. If your league requires that you start two quarterbacks, the same method could be applied, but the specific numbers given in this article are of no use.

• This article ignores strategic pairing of quarterbacks. Handcuffing, for example, is probably relevant but is not addressed here. To name a specific example, Michael Vick and Matt Schaub probably offer more value than an "average" QB12/QB37 combination. Likewise, footballguys.com's Chase Stuart writes an article each year that identifies pairs of mid-round quarterbacks that have complementary schedules. If, to make up an example, QB11 plays easy teams when QB15 plays tough teams and vice versa, then the QB11 / QB15 tandem ought to generate more production this year than a "typical" QB11/QB15 combo.

• The article also ignores whatever advantage (or disadvantage) you might get from playing matchups. If you think you can successfully play matchups week-to-week, then you might be able to get more production out of the QB platoons that don't contain a stud. Thus, the model might be overstating the difference between Manning and a platoon of lesser QBs.

• Remember that this article's purpose is to help you assess the merits of several very broad general draft strategies. It helps to answer the general question "how long should I wait to take a quarterback." It says nothing about which quarterback to take, which is obviously at least as important.

• When I say, for example, that you should expect 306 points out of the QB6/QB19 pair, I'm giving you an expected value. What I am not giving you --- and what I can't give you until I have another decade of data --- is the variance. If QB1/QB24 gives you a very predictable 315 (i.e. you can pretty much count on it being between 300 and 330) while QB6/QB19 gives you a wide variety of numbers having an average of 306 (i.e. it might be 250 or it might be 360), then that might be something you'd want to take into account.

A preview of the sequel...

As I hinted earlier, the cost of those 50 or 75 points you lose by waiting to take your QB can only be put into context if we know how much you stand to gain at other positions by waiting on the quarterback position.

The obvious next step is to do a similar analysis for each of the other positions. But it's going to get complicated. Assuming a 12-team league with four roster slots devoted to running backs, the questions start to look like this. Which is better: RB1/RB16/RB31/RB48 or RB1/RB26/RB31/RB37? There are 194,580 ways to choose four running backs from a list of 48, so the results will obviously be tough to present. But I've got some tricks up my sleeve.

After we know the relative value of each combination of running backs, receivers, and tight ends, then we can start talking about piecing them together. For instance, if you're drafting in the 12th position in a 12-team league and you go RB, RB, WR, QB, WR, then you might estimate that you'd end up with RB10, RB11, WR7, QB3, WR22, or something like that. What is the value of that group compared to what you might be able to get by going WR, WR, RB, RB, RB, QB and ending up with, say, WR2, WR3, RB23, RB24, RB33, QB10?

Stay tuned...