At The Motley Fool, we poke plenty of fun at Wall Street analysts and their endless cycle of upgrades, downgrades, and "initiating coverage at neutral." Today, we'll show you whether those bigwigs actually know what they're talking about. To help, we've enlisted Motley Fool CAPS to track the long-term performance of Wall Street's best and worst.

And speaking of the worst...
Shareholders of semiconductor maker NVIDIA (Nasdaq: NVDA) got quite a shock yesterday. First, their company warned of weaker-than-expected revenues for the current quarter. There's fierce competition over smartphone slots, for one thing, and slack demand for graphics chips in laptops and PCs for another.

Thailand's flooding last year, you see, caused hard-drive supply shortages at Western Digital (NYSE: WDC) and Seagate (Nasdaq: STX). The difficulty of obtaining hard drives is expected to crimp personal-computer sales, and higher prices on the drives available will also sap demand. Result: fewer NVIDIA chips going into PCs that aren't bought. As a result, NVIDIA now sees sales falling about 11% short of consensus estimates in the current quarter.

Soon after this news was out, analyst JMP Securities downgraded NVIDIA shares to underperform (aka sell), citing among other things NVIDIA's "lackluster" record for getting its new Tegra 3 chip into smartphones. Should investors be worried?

Let's go to the tape
Perhaps. You see, in many respects, it's unfair to label JMP Securities a "worst" analyst. For one thing, this banker often gives good advice on the stocks it recommends, getting about 48% of its recommendations right (nearly as accurate as a coin flip). For another thing, when JMP does guess right, it often picks really big winners. The average JMP recommendation, counting both winners and losers, still outperforms the S&P 500 by about 14 percentage points. So as I say, JMP is a pretty good analyst on many things.

Semiconductors is not one of these things.

A record to run from
Over the six years we've been tracking its performance in semiconductors, JMP has racked up a perfectly miserable record of just 33% accuracy on its chip stock recommendations. Quite literally, this analyst has proven itself twice as likely to be wrong on its semi picks as it is to be right. Even more important is JMP's record on NVIDIA itself. Three times in the last four years, JMP has issued affirmative buy-sell advice on NVIDIA. JMP's been wrong each time, with picks that have underperformed the market by 28 percentage points, combined.

And yet, the analyst is sticking to its guns on NVIDIA. Two months ago, as you may recall, JMP downgraded the stock to market perform, citing worries that NVIDIA was losing too many smartphone slots to competing chips from Qualcomm (Nasdaq: QCOM) and Texas Instruments (NYSE: TXN). As the analyst described it, the 85 products that had adopted Tegra paled in comparison to the "hundreds" of design wins notched by Qualcomm and TI. Wins that have gone beyond smartphones to edge out NVIDIA in tablet computers such as the Xoom2, Kindle Fire, and Nook.

JMP's opinion on this score hasn't changed; it's strengthened. So before earnings come out next month, JMP wants to make sure it's on record predicting a disappointment.

Curb your enthusiasm
But here's the thing: We already know NVIDIA is going to disappoint us (to some extent). That's why the stock is so cheap! At 14 times earnings and just 11 times trailing free cash flow, it seems clear that few investors are expecting a blowout quarter from NVIDIA in February. Maybe not even in May. But if NVIDIA delivers anything like the growth that Wall Street expects for it further out, the stock is still crazy-cheap at today's prices.

Consider: Fourteen times earnings is just about the perfect price you'd expect for the 14% growth the Street has NVIDIA pegged for. Back out the company's $2.7 billion cash hoard, and the ex-cash P/E falls below 10 times earnings. Value the stock on free cash flow, and it's cheaper still -- with an enterprise-value-to-free-cash-flow ratio of just 7.8.

Foolish takeaway
In short, if NVIDIA delivers just half the expected long-term growth analysts expect from it, it's fairly priced today. If it does anything better than that, it's a bargain. That's why I've personally -- and publicly -- recommended the stock in my CAPS portfolio (which you can see here, along with 55 other recommendations I've made, and my record on each). That's why I'm convinced JMP is wrong to sell NVIDIA today.

Looking for more ways to make money on the smartphone revolution? Read our new -- and free! – report: " 3 Other Hidden Winners of the iPhone, iPad, and Android Revolution ."