Goutama's final EPS Contest Results (all categories)
Here's a list of winners for all categories of the EPS contest. The contest included 7 categories:
1. EPS estimate
2. K7 (Athlon) units (millions)
3. K7 ASP
4. K6 units (millions)
5. K6 ASP
6. Memory revenues (millions)
7. Closing stock price day after release, i.e., 1/20/00 close
We combined items 2/3 and replaced it by K7 revenues, and items 4/5 were replaced by K6 revenues.
Using the best information from the earnings statement, earnings numbers and the conference call, we came up with the following for the actuals:
The EPS was 0.426056
The K7 revenues were 199.2 million ($249 x 0.8 million)
The K6 revenues were 291.2 million ($56 x 5.2 million)
Memory revenues were 275.0 million
Closing stock price was $38 (the NYSE close)
We combined all categories together by weighting errors by 1/(standard deviation of the guesses). The sum of the absolute values of the error then became a "total error." In non-mathematical terms this means that, to get a good score, K7 revenues, which were all over the map, did not have to be as accurate as memory revenues, which had much less variance. On the theory that guessing EPS was the most important point of the contest and K6 revenues the least important guess, a ranking was also done for which these items are weighted double and half, respectively.
BEST OVERALL ESTIMATES -- ALL 5 CATEGORIES EQUALLY WEIGHTED
Error=average error of each category in standard deviations
Name.....EstimateDate...Thread...Error
1. Goutama 1/18/00 SI 0.50
2. P3killer 1/1/00 Yahoo 0.53
3. nosaert 12/5/99 Yahoo 0.59
4. MYNGA 12/29/99 SI 0.60
5. SI_Mean (average date is 12/10/99) SI 0.64
6. John Petzinger 11/22/99 SI 0.66
...
10. Yahoo_Mean (average date is 12/22/99) Yahoo 0.68
...
15. JC's_Mean (average date is 12/7/99) JC's 0.74
Other than Goutama winning, the most amazing thing here is how accurate the mean estimates from the three threads are. If you used the SI average to predict EPS, Stock Price and the other variables, you'd come awfully close to winning this contest and beating all 20 wall street analysts by a large margin. Note that Goutama's "error" of 0.5 means that his estimates were an average of 1/2 of a standard deviation away from perfect in all the categories.
BEST OVERALL ESTIMATES -- EPS Double-Weighted & K6 Half-Weighted
Error=average error of each category in standard deviations
Name.....EstimateDate...Thread....Error
1. p3killer 1/1/00 Yahoo 0.40
2. Goutama 1/18/00 SI 0.42
3. godefl 12/7/99 Yahoo 0.52
4. George m Lucas Jr 12/29/99 SI 0.55
5. John Petzinger SI 11/22/99 0.55
6. SI_Mean (average date is 12/10/99) SI 0.57
9. YahooMean Yahoo (average data is 12/22/99) 0.66
17.JC'sMean (average date is 12/7/99) JC's 0.79
BEST EPS ESTIMATES
p3killer 1/1/00 Yahoo $0.42
DRBES 11/15/99 SI $0.44
Technically, p3killer beat DRBES because AMD's eps was actually 42.6 cents per share. However, DRBES made his estimate midway through the quarter.
SI with a 0.37 mean estimate beat Yahoo's 0.35. (Note that Elmer's -0.62 estimate was not counted in the SI averages since he said it was a joke.)
BEST STOCK PRICE ESTIMATES
Name....EstimateDate...Thread
ratdogg 11/21/99 JCs
MYNGA 12/29/99 SI
These two guys both nailed it, but the following were all within $1:
Richard Mullin 11/22/99 JCs
godefl 12/7/99 Yahoo
Kash Johal 11/24/99 SI
tezek 11/24/99 SI
p3killer 1/1/00 Yahoo
JC'sMean 12/7/99 JCs - yes JC's thread had the best stock price by far
TechStockBroker 12/24/99 Yahoo
Mark Lucas 12/2/99 SI
All the threads were too optimistic, but I found that if estimates made after 12/31/99 were discarded, the estimated stock price was only high by 0.8! Apparently, as AMD's stock price rose, the new estimates for price went higher and higher. In reality, part of the gain from the coming good news was already reflected in the stock price, causing the later estimates to be too high. (The average estimate made in January was $52)
BEST K7 REVENUE ESTIMATES
fl_tom 11/26/99 JCs 0.8M@$250=$200M
nosaert 12/5/99 Yahoo 0.9M@$260=$234M
actual=0.8M@$249=$199.20M
The SI average was high by nearly $100M and the other threads were even worse. :-(
BEST K6 REVENUE ESTIMATES (actual=$291M)
amd_france 1/14/00 Yahoo $292
GMathiasf 1/14/00 Yahoo $290
JUR 11/21/99 JCs $294
Michael Dakota 12/18/99 SI $295
TechStockBroker 12/24/99 Yahoo $296
YahooMean 12/22/99 Yahoo $285.69
In general, K6 revenue estimates were much closer to the truth than the Athlon estimates.
BEST MEMORY REVENUE ESTIMATES (actual=$275M)
bpowellx 1/12/00 Yahoo $278M
JC'sMean 12/7/99 JCs $270.81
YahooMean 12/22/99 Yahoo $270.43
Estimates were very good here. Fifteen entries had either $270M or $280M (off by 5 million) for their estimate.
How did we come up with the K6 and K7 revenues shown above? Basically, the following information was available:
Since Sanders said that AMD "met their goal" for booked sales, the number 0.8M Athlons was assumed.
Sanders also said that Athlon ASP was "less than $250.
Total CPU revenues (not including chipsets and embedded) were given as "nearly $500M."
CPG revenues were $577M and embedded+chipsets were $80M, therefore K6revs+K7revs were 497M.
Over 5M K6's
6M total CPU's. Above two facts imply ~5.2M K6's
One other thing: K7 revenues were > 50% of increased CPG group sales.
This would be impossible with the known numbers for Q3 and Q4 unless these increased sales also included incremental chipset sales for Q4 -- a reasonable assumption.
Petz |
