Hi all; PTNewell gets touchy about his past predictions. Might as well answer him.
On Forecasting
Motley Fool Rambus thread, #53290, November 12, 2001
elixe (1) chastises me for past forecasts, in doing so digging up my very first calculation posted on YHOO. That post contained simple errors (e.g., the tax rate was listed, but not actually applied). These errors were corrected by myself within hours, before anyone else caught them, and revised calculations posted. This attack by elixe follows the previous pattern where some technical calculations of mine posted on YHOO were revived, even though I had already since posted more sophisticated calculations on TMF (and YHOO). Now I have surely made far more than two errors in message board posting (the format lends itself to such, particularly YHOO). Yet my critics are enamoured of these two posts. elixe, as Hugo before him, resorts to digging up a message off the YHOO board which is more than a year old, AND FOR WHICH I MYSELF POINTED OUT THE ERROR, AND MADE A CORRECTION. Put aside for a moment the obsessiveness of searching out 17-month old YHOO postings. Put aside if you can (I cannot) the clear dishonesty of posting, as elixe just did (1) a very old YHOO message which I corrected myself promptly (before anyone noticed, or at least commented), without mentioning the quick correction. The obvious intent of such crucial omission is to mislead. But even putting that aside (perhaps elixe and even Hugo have been misled by another, one more obsessive, and less honest than themselves, one who posts day and night on the least read of the Rambus boards, and seems inordinately fond of reading me) a peculiar circumstance still exists. These critics attack only on points where I first pointed out the mistake. Are you not capable of challenging me on any point UNTIL AFTER I point it out? Are you not capable of doing better than digging up 17 month old YHOO messages?
Clearly you are not.
[Bilow: The elixe post is here:
boards.fool.com
In it, elixe noted that PTNewell predicted (in December 2000) that Rambus was going to earn (post split) $12.24 per share and be worth $245 per share in 2002. I don't think that the error was a matter of some subtle omission, LOL!!!]
Now if I were prone to such pettiness, I would address something recent, and something posted on this board.
[Bilow: Actually, it's a lot more fun to look at PTNewell's predictions that are old enough that they've had time to be proved wrong.]
Say, elixe's recent claim that PC1066 will have 4 bits simultaneously on a data bus. Since the typical length on a RDRAM board is about 25 cm, and the speed of light is 3x10**10 cm/s, in fact a bit takes less than 25 cm/(3x10**10 cm/s) = 0.8 nS. Thus PC1066 should usually not have more than ONE bit on the bus, and never the 4 that elixe claimed. Such a counter-thrust would be both more relevant and more topical (and of course, elixe has, I believe, not pointed out his own error). In short, if I were inclined to join in such petty attacks, I would at least take the care to make them to the point.
[Bilow: Unfortunately, the good professor blows it again, not once but twice:
(1) The speed of propagation on a circuit board is not the speed of light, but instead about half that.
(2) Each bit travels down the wire twice (during a memory read). They are emitted from the memory, travel to the memory controller, are reflected, and travel back down the wire to the other end, where they are absorbed by the termination.
The combination leads to a design where, yes, 4 bits are simultaneously on the bus.
Why would anyone listen to a loser who is unable to predict the stock price, the earnings, the market share, and doesn't even understand the basic physics of Rambus?]
Still, let us spend a moment thinking about forecasting. Many academic researchers believe it is a waste of time. In the influential book A Random Walk Down Wall Street, , Burton Makiel cites many academic studies showing that Wall Street analyst's predictions are worse than useless. Specifically, the studies show that it is more accurate to simply assume that the earnings for last quarter will be good for the next quarter, or that earnings for last year will persist into the next year, than to use any analysts predictions. A sobering thought.
Nonetheless, we all want to at least predict whether earnings are increasing or decreasing, and at least, whether a company is profitable or unprofitable. So let us see how that debate has played out over the last year in the context of Rambus.
First, the huge collapse of the DRAM industry, which has been far worse than any previous slowdown (unlike, say PC sales themselves, which are in a slump, but far from a historically worse slump) was not predicted by anyone. In fact, every major forecaster predicted an upswing in 2001, not a huge drop. Thus I do not feel guilty than I did not foresee the DRAM collapse, I know of no one who predicted such a collapse before the fact (it is hard to predict the worst collapse ever).
I do feel stupid about taking DataQuest's forecast from about 1.5 years ago of a $76 billion market (and further increases up to $100 billion by 2004) seriously. BusinessWeek has documented how DataQuest made extremely optimistic predictions in virtually every tech area from fiber optics to web hosting, from web site design to cell phone usage. Probably none of those predictions made by DataQuest in 1999 and 2000 will remotely come true. Taking these forecasts seriously is about the only thing I really feel stupid about.
[Bilow: It's well known in the industry that Dataquest is clueless. Heck, I've known this for years, but then again, I'm in the industry. Certainly the losers around here believed in Dataquest back when Dataquest had figures that made Rambus out to be the big winner. I was gloating over Dataquest's failed predictions in August 2000, 4 months before PTNewell was using their figures to predict that Rambus was going to go to $245 per share: #reply-14141872
The problem with listening to predictions from PTNewell is that he doesn't know the first thing about the industry. He doesn't know who to believe and who to ignore. He's a physics professor, not a memory designer.]
On the other hand, as poor as official (and unofficial) forecasting is, it is certainly nice to get the most basic points right. (1) Will the company make money or lose money, (2) If profitable, will the profits rise or fall?
[Bilow: Actually, what is far more useful is the question of will the stock go up or down, and on this PTNewell was very wrong. The second most useful question was what will be the next standard memory type, and on this PTNewell was wrong again. As far as the question of profitability, Rambus is only showing profitability now because they are refusing to book court judgements against them, LOL!!!]
Semico did not predict as large an increase for the DRAM market as DataQuest, but they too predicted an increase, not a calamitous decrease. [Bilow: I guess Semico did a better job.] Moreover, Semico specifically predicted (2):
In her "DRAM Market Direction" address last week, Sherry Garber, senior vice president of Semico Research Corp., Phoenix, estimated the much-heralded but controversial RDRAM would gain 2.6 percent of the total units shipped this year, then trail off to about 2.3 percent of the market in 2001. Rambus RDRAM then declines to 1.1 percent of total units shipped in 2002 and 0.8 percent in 2003.
I believe it is extremely well documented that current RDRAM production (10 million 128 Mb/month from Samsung alone) utterly dwarves RDRAM production from 2000. Thus Semico did not even get the direction right: they predicted RDRAM production declining in 2001, when actually it increased sharply. It is hard to be more wrong than that. Semico managed that apparently impossible trick however: they also predicted DDR sales would surge strongly in 2000. Oops.
[Bilow: Actual RDRAM figures for 2001 are what, around 7.5% of total? So Sherry was off by a factor of three or so. By contrast, PTNewell was off by a factor of 30 on the stock price. By the way, his link to the Electronic News article didn't work, but I'll trust him to have quoted correctly.]
It is also well worth mentioning that many reports appeared in the media from analysts who predicted that Rambus would lose money in 2001, as RDRAM never took off (BusinessWeek had an article from a professional short describing this likelihood, for example). On the messageboards, talk of Rambus bankruptcy was everywhere. Actually, Rambus profits in this last quarter, namely 0.06/share, would have been $0.24/share pre-split, and are thus up sharply from early or mid 2000. Few technology companies can say as much.
[Bilow: But Rambus isn't accounting for the 7 cents per share they lost in a judgement to Infineon. And who knows how much the other lawsuits are going to cost. But what the heck, let's keep on investing in Rambus based on the sunniest thing that can possibly happen.]
Of course Rambus has in particular also been hit by the horrible legal setback in Richmond, which not only put a stop to SDRAM signings, it led the largest SDRAM licensee (Samsung) to renegotiate payments at a lower fixed rate. But the Richmond court rulings outrage me, rather than make me feel stupid. I still hope to see justice done, and thus that outcome reversed. Meanwhile, please note how impressive is Rambus's ability to have such strongly positive earnings despite the meltdown and legal setbacks.
[Bilow: PTNewell wouldn't be outraged if he would look at the situation from a balanced point of view, just like the federal judge, the jury, and the 3-judge appeals panel did. What is he going to say when Rambus loses their next appeal?]
Obviously Rambus skeptics from a year ago were completely unrealistic (more than 100% in error). I know of no long time Rambus skeptic who predicted RDRAM sales still increasing late in 2001. [Bilow: Rambus sales are decreasing, not increasing. This is evident to any fool who tries to buy an RDRAM based computer. The P-IIIs with RDRAM are gone, and the P4s are almost entirely SDRAM now. Plus, it's been shown multiple times that Samsung is no longer increasing their RDRAM production, and Toshiba and NEC are decreasing all DRAM production.] Thus these long time skeptics can heretofore be safely ignored, for their obvious inability to even get the sign right (increasing market share or decreasing). [Bilow: This is sophistry of the worst sort. How can a scientist type stuff like this?] Two years (or more) of being wrong quarter after quarter is quite enough. On the other hand, Rambus shareholders should take pride that their belief in the power of the IP model, and Rambus's patent portfolio in particular, has been vindicated by the ability to weather such a perfect storm (rising legal fees, falling SDRAM royalties, collapse of DRAM industry prices).
But of course, AMD analysts and shareholders could not even get the sign for their company right. [Bilow: Why the change of topic to AMD? The real problem for Rambus now is Intel, not AMD. Intel is the company that will force DDR to be standardized, not AMD.] Not only have AMD's profits not increased. Not only did earnings plummet sharply. The company is now actually losing money. It lost money last quarter, it will lose more this quarter, etc. I am not inclined to take much flack on predictions from any long-time AMD supporter, since clearly none of them even realized that their company was headed negative. That of course implies an error of greater than 100%. [Bilow: PTNewell conveniently fails to calculate the percentage error that his predictions were wrong by. He had predicted an increase in share price of about 500%, but what he got instead was a decrease of around 80%.] (Incidentally, why is it that in digging up my old YHOO messages, these persistently wrong critics do not post my comments on how MU and AVNX, among others, were a good short? [Bilow: Funny that RMBS was a better short than MU.] Never mind, I know the answer.) It is no great accomplishment to have been correctly pessimistic about a technology company over the last year. Micron is another of many companies where the analysts in 2000 could not even get the SIGN of next year's earnings correct. [Bilow: Again this insistence on reducing earnings to "signs". Signs are nearly useless, it's the magnitude of the earnings or losses that are important. Which would you rather have, $1000 - $1 or $1 -$1? If all you cared about was signs you wouldn't be able to distinguish these.] If one uses GAAP accounting, (including the special charges companies like Intel and Cisco have recently taken, and which MU apparently takes every quarter) analysts would have the sign wrong on virtually every major technology company. In that context, optimism about the future of Rambus's earnings in mid-2000 seem more prescient than puerile. At least one would have gotten the correct sign (and thus be less than 100% off), and have been correct that earnings were increasing. [Bilow: Actually, PTNewell was calling for earnings to increase many times. In this he was wrong. The only reason he's getting his error to be wrong by only 100% is because he's ignoring what the status of Rambus was at the beginning of the calculation. He thought that earnings were going to zoom, instead they dived. He got the sign wrong. But who cares about the details of this, the fact is that Rambus was supposed to be immune to a tech downturn because of their IP business. They were supposed to be immune to a loss in RDRAM revenue because of their patents. They were supposed to be immune to a loss in RDRAM market share because of Intel. All these things turned out to be wrong.] Indeed, it would be quite difficult to find any technology stock proponent on any message board anywhere who did as well. It would be even harder to find any critics of any technology company anywhere who did as poorly at predicting the future as Rambus bears have done. As long as one discusses earnings or market share (as opposed to stock price) Rambus bears have arguably been the worst forecasters and have the least successful track record in the recent world of technology stocks. [Bilow: I guess that's why the bears took home the profits.] Show me any bear posting on any Rambus board in early 2000 who predicted that RDRAM would still be outselling DDR by late 2001. [Bilow: It's not, moron.] Show me any long time bear who has not been wildly overoptimistic about DDR (recall the foolish claim that DDR would be the obvious winner by summer 2000, a claim later papered over by silly reference to 'design wins' which had in fact happened in 1999?) [Bilow: It was obvious to all (reasonable forecasters) that DDR had won by the summer of 2000. The clue was when (a) Intel came out with full support for DDR on the desktop, and (b) no other company came out with support for RDRAM.]
boards.fool.com
By the way, my favorite PTNewell prediction was when he said that DDR PC2100 was against the laws of physics:
messages.yahoo.com
How does he spin that one now that PC2700 is selling? Maybe the laws of physics have changed since then.
-- Carl |
